text,label,id
a image super of resolution great low-level vision Single computer as importance task. is,Single image super resolution is of great importance as a low-level computer vision task.,0
approaches achieved networks Recent neural im-pressive performance. with deep have convolutional,Recent approaches with deep convolutional neural networks have achieved im-pressive performance.,1
"hierarchical multi network the different kernel Specifically, multi with learns under fields. representations receptive","Specifically, the network with multi kernel learns multi hierarchical representations under different receptive fields.",2
identified can when be terminated Tracking direction with is high certainty. no,Tracking is terminated when no direction can be identified with high certainty.,3
This placed manually and centerlines correspondence extracted between showed strong markers.,This showed strong correspondence between extracted centerlines and manually placed markers.,4
an of partitioning meaningful image segments. segmentation  process Image  the is into,  Image segmentation is the process of partitioning an image into meaningful segments.,5
algorithm. over approach demonstrates conventional improvement The watershed an,The approach demonstrates an improvement over conventional watershed algorithm.,6
"have vision systems. to made computer to significant attention exploit efforts advance schemes been Recently,","Recently, significant efforts have been made to exploit attention schemes to advance computer vision systems.",7
"target objects For undergoing visual to it often is large track appearance tracking, changes. challenging","For visual tracking, it is often challenging to track target objects undergoing large appearance changes.",8
by tracking selectively paying features. robust facilitate temporal attention visual to maps Attention,Attention maps facilitate visual tracking by selectively paying attention to temporal robust features.,9
the change scene. allow Video systems the Free-Viewpoint of freely the to viewpoints (FVV) viewers,Free-Viewpoint Video (FVV) systems allow the viewers to freely change the viewpoints of the scene.,10
"subjective analyzed. carried test was including and assessment different HRTs Firstly, out a","Firstly, a subjective assessment test including different HRTs was carried out and analyzed.",11
quantify The demonstrate the in and perceived influence results the HRT of quality.,The results demonstrate and quantify the influence of HRT in the perceived quality.,12
light This the simulate and more. costume allow may interaction and dynamically changing,This may allow changing the costume dynamically and simulate light interaction and more.,13
"common the video-mapping, to warping image offline, once, show. done before be Contrary cannot the","Contrary to common video-mapping, the image warping cannot be done once, offline, before the show.",14
"sparsity, yet channel regarding not is available.","regarding channel sparsity, is not yet available.",15
measurements. channel first by of evaluation data The from channel been sounder has verified,The channel sounder has been verified by evaluation of channel data from first measurements.,16
"design residual charge provides prediction. we auxiliary for unit, a information which attention Then","Then we design a attention residual unit, which provides auxiliary information for charge prediction.",17
"addition, SECaps our charges. model In focal imbalanced the loss, problem which relieves of introduces","In addition, our SECaps model introduces focal loss, which relieves the problem of imbalanced charges.",18
superiorities experimental consistently of the model. demonstrate and competitiveness The results proposed our,The experimental results consistently demonstrate the superiorities and competitiveness of our proposed model.,19
develop a successive problems. algorithms of convex that these achieve stationary We point non-convex approximation,We develop successive convex approximation algorithms that achieve a stationary point of these non-convex problems.,20
"To novel, the first present locally for end, user expression. tight this bounds we rate","To this end, we first present novel, locally tight bounds for the user rate expression.",21
complexity and degradation. approximate detectors without reduced also noticeable derive We computation of performance,We also derive approximate detectors of reduced computation complexity and without noticeable performance degradation.,22
"still, challenging it remains a But problem.","But still, it remains a challenging problem.",23
present for we cascaded this paper approach brain In tumor deep segmentation. automatic,In this paper we present deep cascaded approach for automatic brain tumor segmentation.,24
crosstalk in demonstrate We improve and can concentration precision discrimination sensing tasks. that,We demonstrate that crosstalk can improve precision in concentration sensing and discrimination tasks.,25
receptor ubiquity help observed of in crosstalk Our results molecular the rationalize sensing.,Our results help rationalize the observed ubiquity of receptor crosstalk in molecular sensing.,26
"repeated of in composition remove of First, designs we terms composed layers.","First, in terms of composition we remove designs composed of repeated layers.",27
"efficiency achieved. in two better Last, could a cases parameter remove where we designs be","Last, we remove designs in two cases where a better parameter efficiency could be achieved.",28
"four detailed analysis Additionally, designs we scheme. provide in final on efficiency the our","Additionally, we provide detailed efficiency analysis on the final four designs in our scheme.",29
"could to However, terms lead inefficient this an route cost. of traveling in","However, this could lead to an inefficient route in terms of traveling cost.",30
solve proposes two d-relaxed the to rule. approaches with priority research problem Our,Our research proposes two approaches to solve the problem with d-relaxed priority rule.,31
the method. to solution formulation exact proposed construct improve an We the literature mathematical in,We improve the mathematical formulation proposed in the literature to construct an exact solution method.,32
effectiveness Experimental the methods. show of results our,Experimental results show the effectiveness of our methods.,33
"population-level variability models and models. clinically shape Deformetrica Furthermore, relevant ShapeWorks compared capture to SPHARM-PDM","Furthermore, ShapeWorks and Deformetrica shape models capture clinically relevant population-level variability compared to SPHARM-PDM models.",34
backbone networks the for line power as The VLCs. are indoor existing used,The existing indoor power line networks are used as the backbone for VLCs.,35
provided the of analysis. correctness the throughout are simulations Monte verify to the paper Carlo,Monte Carlo simulations are provided throughout the paper to verify the correctness of the analysis.,36
module. into the of policy the auxiliary reward communication for is integrated learning The,The auxiliary reward for communication is integrated into the learning of the policy module.,37
generalize. to easy We effective this reward auxiliary is show that empirically and,We show empirically that this auxiliary reward is effective and easy to generalize.,38
estimation image represents vision. a in single a very exciting challenge from computer Depth,Depth estimation from a single image represents a very exciting challenge in computer vision.,39
"outperforms depth on KITTI methods state-of-the-art the In particular, dataset monocular network our for estimation.","In particular, on the KITTI dataset our network outperforms state-of-the-art methods for monocular depth estimation.",40
quantities approach when approaches works and are inapplicable. rendering Our prices standard latent,Our approach works when quantities and prices are latent rendering standard approaches inapplicable.,41
"the During all boxes from detector. DCR testing, base refines","During testing, DCR refines all boxes from the base detector.",42
and on Experiments results PASCAL bells any show competitive without VOC COCO and whistles.,Experiments show competitive results on PASCAL VOC and COCO without any bells and whistles.,43
"scanners. across existing exhibit MRI reproducibility However, low generally techniques","However, existing techniques generally exhibit low reproducibility across MRI scanners.",44
"Specifically, entire the a global-level, information uses the DSGD grasp. predict % at image to","% Specifically, at the global-level, DSGD uses the entire image information to predict a grasp.",45
"During as grasp selects the output. the most confident inference, % DSGD","% During inference, DSGD selects the most confident grasp as the output.",46
grasp limitations generated overcomes of individual models. candidates hierarchically the from selection This,This selection from hierarchically generated grasp candidates overcomes limitations of the individual models.,47
of terms accuracy. in grasp the state-of-the-art on Cornell DSGD % outperforms dataset methods grasp,% DSGD outperforms state-of-the-art methods on the Cornell grasp dataset in terms of grasp accuracy.,48
"space. instances on cohesion of Importantly, they property separation rely in feature the and among","Importantly, they rely on the property of cohesion and separation among instances in feature space.",49
composed the networks. then output by three The masks semantic are of final segmentation the,The final semantic segmentation masks are then composed by the output of the three networks.,50
slab for algorithm an numbers This factory identification in recognizing paper proposes scenes.,This paper proposes an algorithm for recognizing slab identification numbers in factory scenes.,51
"global boundedness the model. on results of stability, The leads dynamical bifurcations to and analysis","The dynamical analysis leads to results on stability, global boundedness and bifurcations of the model.",52
"Next, ecological control outcomes. predict methods various via we optimal","Next, via optimal control methods we predict various ecological outcomes.",53
sums start-ups. at been AI of money Vast thrown have,Vast sums of money have been thrown at AI start-ups.,54
Integration technique to refer vectorization this Measure Vectorization as PageRank (PMIV). We,We refer to this vectorization technique as PageRank Measure Integration Vectorization (PMIV).,55
"the great progress challenging achieved,it's match still scenario,i.e. multimodal in Despite the to","Despite the great progress achieved,it's still challenging to match in the multimodal scenario,i.e.",56
that shown dynamic It recently filters SISO DFE has (i.e. with MMSE been,It has been recently shown that SISO MMSE DFE with dynamic filters (i.e.,57
of learning various and plays in applications. very important visual roles vision images qualities Enhancing,Enhancing visual qualities of images plays very important roles in various vision and learning applications.,58
"the the frameworks we task review and first Specifically, each relevant general for introduce variants.","Specifically, we first review the general frameworks for each task and introduce the relevant variants.",59
limitations are method The and also advantages each of discussed.,The advantages and limitations of each method are also discussed.,60
"evaluation comprehensive we comparison metrics. present different Moreover, and benchmark datasets of a","Moreover, we present a comprehensive comparison of different benchmark datasets and evaluation metrics.",61
"Finally, explore challenges and of parsing. the trends semantic we image future","Finally, we explore the future trends and challenges of semantic image parsing.",62
and a latter an The iterative is method solution. former is closed-form the,The former is an iterative method and the latter is a closed-form solution.,63
"the latter low-complexity. is the former, of Compared to","Compared to the former, the latter is of low-complexity.",64
in methods of accurately The cannot condition traditional aerators abnormal detect time.,The traditional methods cannot accurately detect abnormal condition of aerators in time.,65
"feature learning machine classifier. to the we Finally, algorithms build use","Finally, we use machine learning algorithms to build the feature classifier.",66
non-electro-neutral microcirculation to the general lens in a We introduce study of model eyes.,We introduce a general non-electro-neutral model to study the microcirculation in lens of eyes.,67
"main full In captures model. simplified of addition, the our features model the","In addition, our simplified model captures the main features of the full model.",68
Set segmentation. a has been used (VLS) Variational method in medical Level widely,Variational Level Set (VLS) has been a widely used method in medical segmentation.,69
layer at scales. brain learns Convolutional visual tumor different of representation,Convolutional layer learns visual representation of brain tumor at different scales.,70
Layer drives towards the brain Level-Set tumor. the contour,Level-Set Layer drives the contour towards the brain tumor.,71
This input in the next step. layer LevelSet as for serves turn in the,This in turn serves as input for the LevelSet layer in the next step.,72
"end-to-end brain Additionally, segmentation DRLS state-of-the-art a on tumors. system achieves fully","Additionally, a fully end-to-end system DRLS achieves state-of-the-art segmentation on brain tumors.",73
block-level discuss techniques. precoding analog also We,We also discuss block-level analog precoding techniques.,74
by is Neural initialized Networks that Convolutional endcaps. the algorithm detects DLO's a The,The algorithm is initialized by a Convolutional Neural Networks that detects the DLO's endcaps.,75
a of content task generate of to The image. textual description is an the,The task is to generate a textual description of the content of an image.,76
current studies stimuli. saliency and high-quality of modeling gaze on human have Most used,Most of current studies on human gaze and saliency modeling have used high-quality stimuli.,77
"motion variations lighting distortion include and rotation. Some types blur,","Some distortion types include motion blur, lighting variations and rotation.",78
"the we effectiveness serving transformations. distortions when of augmentation data investigate Finally, different as","Finally, we investigate the effectiveness of different distortions when serving as data augmentation transformations.",79
Deep single paper proposes for Bi-Dense image Networks (DBDN) This super-resolution.,This paper proposes Deep Bi-Dense Networks (DBDN) for single image super-resolution.,80
dense approach inter-block Our novel intra-block extends including previous connections. dense connection by approaches,Our approach extends previous intra-block dense connection approaches by including novel inter-block dense connections.,81
"encompassing present paper argumentation formalism In AFRA, frameworks. within unlimited we recursive attacks a this","In this paper we present AFRA, a formalism encompassing unlimited recursive attacks within argumentation frameworks.",82
"beamformer designing ADAF the Then, appropriately by the vectors. can adaptive be acquired","Then, the adaptive beamformer can be acquired by appropriately designing the ADAF vectors.",83
point/vertex meshes produce clouds uncorrelated with correspondences. or Most temporally solutions existing unknown point,Most existing solutions produce temporally uncorrelated point clouds or meshes with unknown point/vertex correspondences.,84
compressing ultra-large sizes. and each frame to still ineffective Individually yields is data,Individually compressing each frame is ineffective and still yields to ultra-large data sizes.,85
technique learning-based then We descriptors learn a to on feature pixel-wise PDMs. develop,We then develop a learning-based technique to learn pixel-wise feature descriptors on PDMs.,86
results network compression. are fed an autoencoder-based for into The,The results are fed into an autoencoder-based network for compression.,87
different is of this vehicle. Note the problem the speed predicting from actual,Note this problem is different from predicting the actual speed of the vehicle.,88
apparent highlight deep current These learning. weaknesses limitations of,These apparent weaknesses highlight current limitations of deep learning.,89
has long key task of been theorem a intelligence. Automated proving artificial,Automated theorem proving has long been a key task of artificial intelligence.,90
form inquiry. scientific the bedrock rigorous of Proofs,Proofs form the bedrock of rigorous scientific inquiry.,91
proving little logics. non-classical to attention for theorem automated Very been has paid,Very little attention has been paid to automated theorem proving for non-classical logics.,92
"This to unfortunate, truly many is desire reasons as are there proofs non-classical classical. over","This is truly unfortunate, as there are many reasons to desire non-classical proofs over classical.",93
a neural heterogeneous sensitivity. work that of nodes with consists presents network This,This work presents a neural network that consists of nodes with heterogeneous sensitivity.,94
"Morever, for the analyze methods machine conventional image we systematically denoising. learning","Morever, we systematically analyze the conventional machine learning methods for image denoising.",95
"learning directions for denoising. technologies we in deep the image out Finally, research some point","Finally, we point out some research directions for the deep learning technologies in image denoising.",96
able an undetectable We previously that attack FDI method to derive attacks. detection is detect,We derive an attack detection method that is able to detect previously undetectable FDI attacks.,97
on processing This graph originating (GSP). is method from concepts based signal,This method is based on concepts originating from graph signal processing (GSP).,98
unsupervised such for purposes. an We method propose,We propose an unsupervised method for such purposes.,99
"Visually, the major patterns of depict activities. flows","Visually, the patterns depict flows of major activities.",100
"has its pattern Temporally, each cycles. appearance-disappearance unique own","Temporally, each pattern has its own unique appearance-disappearance cycles.",101
"To method. also propose compute compact we representations, sampling a pattern hybrid","To compute compact pattern representations, we also propose a hybrid sampling method.",102
"better previously. difficult anomalies our Also, data detect that method to were detects fits and","Also, our method fits data better and detects anomalies that were difficult to detect previously.",103
available on is project Our dataset page. the publicly,Our dataset is publicly available on the project page.,104
methods which pre-processing also real-time We performance. propose ensure,We also propose pre-processing methods which ensure real-time performance.,105
"large in occlusion. and The noise, background shadows, approach changes, well viewpoint performs","The approach performs well in large viewpoint changes, background noise, shadows, and occlusion.",106
shown is efficacy The system datasets. crack of standard on road the,The efficacy of the system is shown on standard road crack datasets.,107
of The the extended show model. our and experiments extensibility universality,The extended experiments show the universality and extensibility of our model.,108
the violated. are voltage not and The limits power constraints ensure that line that grids,The constraints that ensure that the grids voltage and line power limits are not violated.,109
proposed reported effectiveness in the of this confirm study the approach. Simulations,Simulations reported in this study confirm the effectiveness of the proposed approach.,110
is preprocessing application of precise step important an the for imagery. satellite optical Cloud detection,Cloud detection is an important preprocessing step for the precise application of optical satellite imagery.,111
been cloud dataset made validation global detection has online. available high-resolution established Our,Our established global high-resolution cloud detection validation dataset has been made available online.,112
"flow the Firstly, model system's transaction analyzed. and are interference block-chain-based","Firstly, the system's interference model and block-chain-based transaction flow are analyzed.",113
"simulation performance by of is algorithm Finally, of different the the verified parameters. proposed system","Finally, the performance of the proposed algorithm is verified by simulation of different system parameters.",114
"demonstrated. unsupervised arithmetic shape and applications, learning also Several operations, as feature such important are","Several important applications, such as unsupervised feature learning and shape arithmetic operations, are also demonstrated.",115
"is thus demanding this prevents extremely real-time and strategy However, inference. computationally","However, this strategy is extremely computationally demanding and thus prevents real-time inference.",116
afs. obtain we concerning finitary results additional the case Moreover of,Moreover we obtain additional results concerning the case of finitary afs.,117
datasets. multiple proposed sketch-photo evaluate network We conducting by experiments baseline the extensive on,We evaluate the proposed network by conducting extensive experiments on multiple baseline sketch-photo datasets.,118
are some and these defuzzification operations of operations. used Aggregation often,Aggregation and defuzzification operations are some of these often used operations.,119
makers produce defuzzification independent decision operators results and aggregation Many the to strategy.,Many aggregation and defuzzification operators produce results independent to the decision makers strategy.,120
"it the to value, However, WABL calculate determine to in is order integrals. some necessary","However, in order to determine the WABL value, it is necessary to calculate some integrals.",121
results the theoretical examples have been Computational explaining performed.,Computational examples explaining the theoretical results have been performed.,122
treatment stroke role a and crucial diagnosis Perfusion plays acute in imaging making. decision,Perfusion imaging plays a crucial role in acute stroke diagnosis and treatment decision making.,123
our approaches. dataset Experimental validate results evaluated our on,Experimental results evaluated on our dataset validate our approaches.,124
current Music Optical our Recognition overview of (OMR) gives This paper research. an,This paper gives an overview of our current Optical Music Recognition (OMR) research.,125
taking We are some both additional of to steps currently them. improve,We are currently taking some additional steps to improve both of them.,126
of graph-based super-voxels an perform clustering a exhaustive these clustering using method. We then,We then perform an exhaustive clustering of these super-voxels using a graph-based clustering method.,127
our to We implement explore clustering intuitive approach. using an volumes user-interface interactively,We implement an intuitive user-interface to interactively explore volumes using our clustering approach.,128
"effectiveness modalities. the multiple real-world datasets of Finally, on we framework different show of our","Finally, we show the effectiveness of our framework on multiple real-world datasets of different modalities.",129
as attributes known Such biometrics. are soft,Such attributes are known as soft biometrics.,130
and gap video. between semantic bridge They the surveillance human description retrieval in person,They bridge the semantic gap between human description and person retrieval in surveillance video.,131
pixel-wise The R-CNN approach Mask segmentation. proposed for uses person,The proposed approach uses Mask R-CNN for pixel-wise person segmentation.,132
the around boundary background precise removes person. clutter It and provides,It removes background clutter and provides precise boundary around the person.,133
challenging accuracy using person the semantic It conditions. query in for good achieves retrieval,It achieves good accuracy for person retrieval using the semantic query in challenging conditions.,134
simulation and CSI the behaviors of The are that also show results performance CDI different.,The simulation results also show that the performance behaviors of CDI and CSI are different.,135
"However, AN with Willie knows CSI, monotonically probability when decreases power. outage the","However, when Willie knows CSI, the outage probability monotonically decreases with AN power.",136
chapter Encyclopedia in Computational appear This Neuroscience. will of,This chapter will appear in Encyclopedia of Computational Neuroscience.,137
datasets proposed available two SBU evaluated method publicly UCF. of and is on The,The proposed method is evaluated on two publicly available datasets of SBU and UCF.,138
"represents a African of of dipteran organisms. one Drosophila model the origin, melanogaster, small best-studied","Drosophila melanogaster, a small dipteran of African origin, represents one of the best-studied model organisms.",139
realistic into challenging is a a a image image Transforming task. RGB infrared thermal,Transforming a thermal infrared image into a realistic RGB image is a challenging task.,140
learning to gap. paper this deep a bridge In we propose this method,In this paper we propose a deep learning method to bridge this gap.,141
transformation propose the We generator coarse-to-fine preserves that the a using mapping details. learning,We propose learning the transformation mapping using a coarse-to-fine generator that preserves the details.,142
approach experiments significantly that existing and approaches. outperforms qualitative our demonstrate Quantitative,Quantitative and qualitative experiments demonstrate that our approach significantly outperforms existing approaches.,143
Complex a sensors. number continuously systems of large are by monitored industrial heterogeneous,Complex industrial systems are continuously monitored by a large number of heterogeneous sensors.,144
"fault. Subsequently, a a on is of evaluated power real approach the case plant study","Subsequently, the approach is evaluated on a real case study of a power plant fault.",145
"prove Of correlation or causation does imply association. not course,","Of course, correlation does not imply causation or prove association.",146
here observation. is presented simply an as It,It is simply presented here as an observation.,147
"estimate non-stationary instantaneous we an this locally, Applying frequency a procedure of signal. obtain multicomponent","Applying this procedure locally, we obtain an instantaneous frequency estimate of a non-stationary multicomponent signal.",148
is simulation. The proposed illustrated by method,The proposed method is illustrated by simulation.,149
to billions The tens provide sensor to nodes. of aims connectivity mMTC,The mMTC aims to provide connectivity to tens of billions sensor nodes.,150
is broadband Denoising a systems. non--stationary communication challenging of in problem signals,Denoising of broadband non--stationary signals is a challenging problem in communication systems.,151
existing connections auxiliary to small layers. FBS introduces convolutional,FBS introduces small auxiliary connections to existing convolutional layers.,152
become has Biorefinery serious years. recent development in sector sustainable for a and cleaner dispute,Biorefinery sector has become a serious dispute for cleaner and sustainable development in recent years.,153
The investigated. hemicellulose concentration of solubilization enhancer on each systematically effect was type and,The type and concentration effect of each enhancer on hemicellulose solubilization was systematically investigated.,154
"by are visual many examination applications, in large-scale collected images for However, not human.","However, in many applications, large-scale images are collected not for visual examination by human.",155
users. PCC-OFDM band frequency required between OFDMA different For is no guard,For PCC-OFDM no frequency guard band is required between different OFDMA users.,156
Animal exhibit interactions. consequence emergent a of local are properties groups that,Animal groups exhibit emergent properties that are a consequence of local interactions.,157
method transition of first system-size The on Kampen's based is expansion of van rates. construction,The first method of construction is based on van Kampen's system-size expansion of transition rates.,158
method Gillespie's Langevin second employs chemical equations. The,The second method employs Gillespie's chemical Langevin equations.,159
"multiplicative, noise. cases, In or mesoscopic derived SDEs state-dependent, the both have","In both cases, the derived mesoscopic SDEs have multiplicative, or state-dependent, noise.",160
estimation. a tree and The procedure commonly is network heuristic spanning in inference adopted,The spanning tree heuristic is a commonly adopted procedure in network inference and estimation.,161
"an intractable dense a in However, of are graph. there trees spanning number","However, there are an intractable number of spanning trees in a dense graph.",162
"address vision Deep networks many problems, used including neural computer video prediction. are convolutional to","Deep convolutional neural networks are used to address many computer vision problems, including video prediction.",163
"modeling spatially neural though, from are Convolutional prevents invariant, networks which location-dependent them patterns.","Convolutional neural networks are spatially invariant, though, which prevents them from modeling location-dependent patterns.",164
"propose convolutional this location-biased work, overcome to the limitation. layers In this authors","In this work, the authors propose location-biased convolutional layers to overcome this limitation.",165
crucial The that task the features location-dependent prediction. results of is indicate encoding for video,The results indicate that encoding location-dependent features is crucial for the task of video prediction.,166
outperform significantly methods proposed models. spatially invariant Our,Our proposed methods significantly outperform spatially invariant models.,167
constrained quadratic formulate inhomogeneous problem problem. We convex programming minimization a linear TV as the,We formulate the inhomogeneous TV minimization problem as a convex quadratic constrained linear programming problem.,168
methods. We TF the draw inspiration from existing our,We draw our inspiration from the existing TF methods.,169
datasets. be pre-trained unlabelled compression The surplus part can using,The compression part can be pre-trained using surplus unlabelled datasets.,170
is in that more effective to DPNN state-of-the-art baselines. existing plausible and show We comparison,We show that DPNN is more effective in comparison to existing state-of-the-art and plausible baselines.,171
We provide new the evidence about that UMT. develop applications two,We develop two applications that provide new evidence about the UMT.,172
"remote face proposed HR methods for estimation Recently, videos. have from been","Recently, methods have been proposed for remote HR estimation from face videos.",173
the database public like the into domain. would put to We VIPL-HR,We would like to put the VIPL-HR database into the public domain.,174
Turing Universal VonNeumann The computers. computers model Machine (TM) --- is a general-purpose for,The Universal Turing Machine (TM) is a model for VonNeumann computers --- general-purpose computers.,175
whether the is same. unknown a can It machine accomplish,It is unknown whether a machine can accomplish the same.,176
how can Developmental shows This (DN) this. theoretical work accomplish Network the,This theoretical work shows how the Developmental Network (DN) can accomplish this.,177
"Master error is of controller central any free or A convolution, Map, back-propagation). DN (e.g.,","A DN is free of any central controller (e.g., Master Map, convolution, or error back-propagation).",178
at shelves pick station. multiple Usually are the there,Usually there are multiple shelves at the pick station.,179
"at a case, head. the with queue picker build In they its this","In this case, they build a queue with the picker at its head.",180
"there answering it. before to because It tough are many question, uncertainties is consider a","It is a tough question, because there are many uncertainties to consider before answering it.",181
"made decision subsequent to the the answer each influences ones. question Moreover,","Moreover, each decision made to answer the question influences the subsequent ones.",182
answer to question goal is properly. the The paper of this,The goal of this paper is to answer the question properly.,183
problem formulate We Repositioning the Pod call model. a and deterministic this Problem,We call this problem the Pod Repositioning Problem and formulate a deterministic model.,184
to reasoners. these essential techniques These are speed up,These techniques are essential to speed up these reasoners.,185
based on choices. Many the of are heuristic optimization techniques,Many of the optimization techniques are based on heuristic choices.,186
heuristic makes these Optimal techniques selection effective. more,Optimal heuristic selection makes these techniques more effective.,187
learning binary to a our SVM use build technique We model.,We use a binary SVM technique to build our learning model.,188
The to order up choose help OWL sets model can speed ontology-specific to reasoning.,The model can help to choose ontology-specific order sets to speed up OWL reasoning.,189
master. Doodling can and learn and common skill intelligent people useful that is a,Doodling is a useful and common intelligent skill that people can learn and master.,190
Numerous images. deteriorations could lead factors medical to partial of,Numerous factors could lead to partial deteriorations of medical images.,191
"to lead perturbations metallic scans. in example, For implants localized will MRI","For example, metallic implants will lead to localized perturbations in MRI scans.",192
"of Generative Networks we propose the inpainting In work, Adversarial (GANs). medical images this via","In this work, we propose the inpainting of medical images via Generative Adversarial Networks (GANs).",193
"and provided. asymptotic Under dependency consistency normality results assumptions, are estimation mild weak on","Under mild weak dependency assumptions, results on estimation consistency and asymptotic normality are provided.",194
"of the DeepNSM results effectiveness methods. our validate state-of-the-art Then, model, exceeding the experimental the","Then, the experimental results validate the effectiveness of our DeepNSM model, exceeding the state-of-the-art methods.",195
criterion the to AT. We for selection OAM impact a of modes minimize propose,We propose a selection criterion for OAM modes to minimize the impact of AT.,196
state-of-the-art methods independently. Current (SoTA) the tasks two treat,Current state-of-the-art (SoTA) methods treat the two tasks independently.,197
while using object moving be modeled could flow. optical easily,while object moving could be easily modeled using optical flow.,198
"as paper, the propose to address a In i.e. two we whole, this tasks","In this paper, we propose to address the two tasks as a whole, i.e.",199
"call Counts++"" as method or We ""EPC++"". Pixel ""Every our","We call our method as ""Every Pixel Counts++"" or ""EPC++"".",200
a Wi-Fi as of medium communication is in chosen HAN. the,Wi-Fi is chosen as a medium of communication in the HAN.,201
incorporates load Python management script. written in The is LMU algorithm which,The LMU incorporates load management algorithm which is written in Python script.,202
"(RSIM). image propose remote we this, unsupervised matching feature sensing for Resolving an encoder-decoder","Resolving this, we propose an unsupervised encoder-decoder feature for remote sensing image matching (RSIM).",203
image been sensing available validated publicly remote two benchmark Results have with datasets.,Results have been validated with two publicly available benchmark remote sensing image datasets.,204
validation approach across of was imaged all Extensive the datasets. conducted,Extensive validation of the approach was conducted across all imaged datasets.,205
"standard Hematoxylin and gold (H&E) staining. pathologists used Eosin Historically, by has been as a","Historically, Hematoxylin and Eosin (H&E) has been used by pathologists as a gold standard staining.",206
tissue also segment. of Virtual same staining in-silico the generate multiplexing on could different stains,Virtual staining could also generate in-silico multiplexing of different stains on the same tissue segment.,207
for critical in karyotyping is diagnosis. Chromosome abnormality classification,Chromosome classification is critical for karyotyping in abnormality diagnosis.,208
global-scale network network one (L-Net). (G-Net) approach local-scale of The one and consists,The approach consists of one global-scale network (G-Net) and one local-scale network (L-Net).,209
learn features. and The first is local both to stage global,The first stage is to learn both global and local features.,210
extract via global and local G-Net. detect We features finer regions the,We extract global features and detect finer local regions via the G-Net.,211
learning utilized extraction. multi-task are high-level promote Residual learning strategies and to feature,Residual learning and multi-task learning strategies are utilized to promote high-level feature extraction.,212
been diagnosis. method karyotype The has to practical assist applied proposed,The proposed method has been applied to assist practical karyotype diagnosis.,213
"resource-intensive deep theory However, developed. is learning yet the is well not and","However, deep learning is resource-intensive and the theory is not yet well developed.",214
"small algorithm be the of simple Q-learning might classical still choice. For games, table-based","For small games, simple classical table-based Q-learning might still be the algorithm of choice.",215
good learning for General testbed reinforcement research Game Playing AGI. to provides (GGP) a,General Game Playing (GGP) provides a good testbed for reinforcement learning to research AGI.,216
"dynamic $\epsilon$ a strategy, For the the propose algorithm. we first enhancement, $\epsilon$-greedy","For the $\epsilon$-greedy strategy, we propose a first enhancement, the dynamic $\epsilon$ algorithm.",217
classical of Both performance enhancements Q-learning. the improve,Both enhancements improve the performance of classical Q-learning.,218
are learning used learning representations. deep models unsupervised for Autoencoders,Autoencoders are unsupervised deep learning models used for learning representations.,219
"publicly Code, available. are and models data","Code, data and models are publicly available.",220
decades despite popular misuse misinterpretation. and remains Null about of hypothesis significance testing concern,Null hypothesis significance testing remains popular despite decades of concern about misuse and misinterpretation.,221
is or correlation see or clearly between-group difference we can a negative). positive that whether,whether we can clearly see that a correlation or between-group difference is positive or negative).,222
in enhance could change communication. simple semantic substantially statistical This clarity,This simple semantic change could substantially enhance clarity in statistical communication.,223
"the difficult the becomes information However, gradually as deepens. is training increasingly network and weakened","However, the information is gradually weakened and training becomes increasingly difficult as the network deepens.",224
"the we branches. information to facilitate also merge-and-run mapping Besides, integration between different adopt","Besides, we also adopt the merge-and-run mapping to facilitate information integration between different branches.",225
to placed challenge. address well be is AI this able to Planning,AI Planning is well placed to be able to address this challenge.,226
The in XAI-Plan framework. the new methodology is implemented,The methodology is implemented in the new XAI-Plan framework.,227
SynSeg-Net advances enabled networks recent is of and cycle generative (CycleGAN) the by DCNN. adversarial,SynSeg-Net is enabled by the recent advances of cycle generative adversarial networks (CycleGAN) and DCNN.,228
performance stage end-to-end approach to methods. The proposed two superior achieved,The proposed end-to-end approach achieved superior performance to two stage methods.,229
transmitters environments. precise communication receivers and between Achieving particularly synchronisation challenging diffusive molecular is in,Achieving precise synchronisation between transmitters and receivers is particularly challenging in diffusive molecular communication environments.,230
the protocol time adaptive search algorithm. proposes paper synchronization a by inspired Newton novel This,This paper proposes a novel time synchronization protocol inspired by the adaptive Newton search algorithm.,231
"is The and robust to implement. simple protocol lightweight,","The protocol is lightweight, robust and simple to implement.",232
the will be by performance with cost. learning constantly enabled hardware Embedded deep low improving,Embedded deep learning will be enabled by the constantly improving hardware performance with low cost.,233
larger possible clinical review in methods systems. computation We briefly will,We will briefly review possible computation methods in larger clinical systems.,234
very networks can classification. Neural perform multidimensional nonlinear effectively,Neural networks can very effectively perform multidimensional nonlinear classification.,235
"electronic lifetimes However, capacitive to bandwidth carrier limitations networks due from significant suffer coupling. and","However, electronic networks suffer from significant bandwidth limitations due to carrier lifetimes and capacitive coupling.",236
These resonators compromising optical microring superior fan-in nonlinearity without and provide bandwidth. both,These optical microring resonators provide both nonlinearity and superior fan-in without compromising bandwidth.,237
Simulations the modes these determined waveguide that optimal geometry validated. were using for experimentally,Simulations that determined the optimal waveguide geometry for using these modes were experimentally validated.,238
face-blurred on framework the both CycleGAN We and face-removed images. applied,We applied the CycleGAN framework on both face-blurred and face-removed images.,239
described as These will be natural transformations. suitable relationsips,These relationsips will be described as suitable natural transformations.,240
of cancer for development the systems. Our is promising screening cervical method automation-assisted,Our method is promising for the development of automation-assisted cervical cancer screening systems.,241
motion from the noise signal. PPG computed reduced is heart and artifacts rate The then,The heart rate is then computed from the noise and motion artifacts reduced PPG signal.,242
"is rate Finally, considering tracking algorithm estimates. neighboring a heart proposed","Finally, a heart rate tracking algorithm is proposed considering neighboring estimates.",243
of performance been on has database. algorithm The the tested available PPG SPECMAR publicly,The performance of the SPECMAR algorithm has been tested on publicly available PPG database.,244
"complexity, than faster is low Due the computational comparing to method methods. the","Due to low computational complexity, the method is faster than the comparing methods.",245
are Visual data and text  information data of at multiple granularities. composed ,  Visual data and text data are composed of information at multiple granularities.,246
"to when learned the embeddings of downstream Furthermore, applied examine we our tasks. effectiveness","Furthermore, we examine the effectiveness of our learned embeddings when applied to downstream tasks.",247
zero-shot its action recognition show video in utility We captioning. and,We show its utility in zero-shot action recognition and video captioning.,248
grid discovered ago a decade cortex of were entorhinal the cells rat. About in medial,About a decade ago grid cells were discovered in the medial entorhinal cortex of rat.,249
demonstrated have in (SR). superior performance networks neural (CNNs) super-resolution Convolutional,Convolutional neural networks (CNNs) have demonstrated superior performance in super-resolution (SR).,250
"presents recursive paper issues, these To a this unit. address novel","To address these issues, this paper presents a novel recursive unit.",251
"and rather features, deep-level only are multi-level extracted features, than Then, fused. the","Then, the multi-level features, rather than only deep-level features, are extracted and fused.",252
"to maps features high-level we First, use the roughly as class-specific classes. locate localization","First, we use high-level features as class-specific localization maps to roughly locate the classes.",253
"networks. approach propose efficiently from we novel this work, In a filters to remove convolutional","In this work, we propose a novel approach to efficiently remove filters from convolutional networks.",254
Projection This Variable computed using and is Least (PLS) Squares (VIP). Partial relationship Importance in,This relationship is computed using Partial Least Squares (PLS) and Variable Importance in Projection (VIP).,255
translation via encode convolution Networks Neural (CNNs) the traditionally operation. Convolutional equivariance,Convolutional Neural Networks (CNNs) traditionally encode translation equivariance via the convolution operation.,256
structured populations dynamics in explored in has past Evolutionary extensively been decades. game,Evolutionary game dynamics in structured populations has been extensively explored in past decades.,257
the ties of social together resulting interactants. with behaviors payoffs determine Strategic,Strategic behaviors together with social ties determine the resulting payoffs of interactants.,258
"evaluation. Keywords: fiber, textile processing, image textile","Keywords: textile fiber, image processing, textile evaluation.",259
"Consequently, focus states. the transitions on is between learning the","Consequently, the focus is on learning the transitions between states.",260
"paper, we assumptions. this drop In such","In this paper, we drop such assumptions.",261
new the by actions. experiences executing learned,the new experiences learned by executing actions.,262
linear for We fields on meshes. present directional a scheme subdivision tangent face-based novel triangle,We present a novel linear subdivision scheme for face-based tangent directional fields on triangle meshes.,263
the diagnosis melanoma systems highly can of and recognition. The efficiency computer-aided improve (CAD) reliability,The computer-aided diagnosis (CAD) systems can highly improve the reliability and efficiency of melanoma recognition.,264
to the apply boost performance approaches We of MS-UNet. two,We apply two approaches to boost the performance of MS-UNet.,265
is few One world. around that in the just exist they countries a,One is that they exist in a just few countries around the world.,266
"a for railways critical such these However, part are low their infrastructure. of frequency countries","However, for these countries such low frequency railways are a critical part of their infrastructure.",267
the second catenary is section. example is The and single-fed section doubly-fed first a catenary,The first example is a single-fed catenary section and the second is doubly-fed catenary section.,268
critical-but-missing a common Artificial of (AI). Machine been has sense Intelligence component long,Machine common sense has long been a critical-but-missing component of Artificial Intelligence (AI).,269
must program Developers carefully every situation. or for train systems,Developers must carefully train or program systems for every situation.,270
"Machine broad, sense AI. remains potentially a in unbounded common problem","Machine common sense remains a broad, potentially unbounded problem in AI.",271
We strategy propose machine neuroanatomical learning novel variation. studying for a shape,We propose a novel machine learning strategy for studying neuroanatomical shape variation.,272
"descriptor invariant scaling. and The affine shape is shifts, to rotations learned including transformations,","The learned shape descriptor is invariant to affine transformations, including shifts, rotations and scaling.",273
"shape. location high problem and ischemic challenging of to due However, variability strokes' remains this","However, this problem remains challenging due to high variability of ischemic strokes' location and shape.",274
variability sizes. This the handling in lesion helps,This helps handling the variability in lesion sizes.,275
industry. to is a maintained weaving be important Quality one things of the in,Quality is one of the important things to be maintained in a weaving industry.,276
implementation fabric. processing chapter on the woven will This of image explain techniques and computational,This chapter will explain the implementation of image and computational processing techniques on woven fabric.,277
"image processing, Keywords: identification, yarn, textiles. digital evaluation and","Keywords: yarn, digital image processing, evaluation and identification, textiles.",278
static deep-learning Activity both otherwise---uses components. videos pre-computed a motion setting---or recognition in and in,Activity recognition in videos in a deep-learning setting---or otherwise---uses both static and pre-computed motion components.,279
decomposition based The method control of Cholesky contributions. the allows,The Cholesky decomposition based method allows the control of contributions.,280
"interpolation. B-spline extending on is As proof of shift-invariant concept, the put conventional focus","As proof of concept, the focus is put on extending conventional shift-invariant B-spline interpolation.",281
This constructing stability satisfies domain-informed is done that generating basis properties. by a,This is done by constructing a domain-informed generating basis that satisfies stability properties.,282
(i) where hierarchical class-level tree we are construct merged recursively. a classes neighboring,(i) we construct a hierarchical class-level tree where neighboring classes are merged recursively.,283
distribution structure whole hierarchical the database. intrinsic data The over naturally captures the,The hierarchical structure naturally captures the intrinsic data distribution over the whole database.,284
context. global meaningful to allows samples of This guide hard select automatically it with the,This allows it to automatically select meaningful hard samples with the guide of global context.,285
"benchmarks, achieves state-of-the-art a number of iterations. on fewer learning new It performance with much","It achieves new state-of-the-art performance on a number of benchmarks, with much fewer learning iterations.",286
algorithm in camera. by captured frame points locates interest of initially the a stereo The,The algorithm initially locates points of interest in a stereo frame captured by the camera.,287
the algorithm Its very for of the management important SLAM. functioning Fast proper is,Its management is very important for the proper functioning of the algorithm Fast SLAM.,288
The results be satisfactory. found are to,The results are found to be satisfactory.,289
(ECG). Effective the of detection an important in is arrhythmia task electrocardiogram of remote monitoring,Effective detection of arrhythmia is an important task in the remote monitoring of electrocardiogram (ECG).,290
"applied Moreover, model. optimize triplet is loss function hard to batch the the","Moreover, the batch hard triplet loss function is applied to optimize the model.",291
a introduced conceptual The of generalization is a as of linkage important notion hyperlink.,The important notion of conceptual linkage is introduced as a generalization of a hyperlink.,292
virtual We transforming problem one. these the (LOS) Line-Of-Sight lift into by a limitations NLOS,We lift these limitations by transforming the NLOS problem into a virtual Line-Of-Sight (LOS) one.,293
light these Our cases method developing explicitly without handles challenging model. transport a,Our method handles these challenging cases without explicitly developing a light transport model.,294
from the is the of same segmenting task images. multiple objects co-segmentation Object,Object co-segmentation is the task of segmenting the same objects from multiple images.,295
"are robust are As to aging features then identity-related for AIFR. result, used a that","As a result, identity-related features that are robust to aging are then used for AIFR.",296
a their to surrounding robots paradigm novel propose teach learning about environment. teacher-student We,We propose a novel teacher-student learning paradigm to teach robots about their surrounding environment.,297
"network ""student"" provides ""teacher"" motion appearance appearance and A two-stream network. an pseudo-labels adapt to","A two-stream motion and appearance ""teacher"" network provides pseudo-labels to adapt an appearance ""student"" network.",298
"IVOS of tasks. videos manipulation contains Our teaching objects, different dataset and","Our IVOS dataset contains teaching videos of different objects, and manipulation tasks.",299
ZIP codes. by the respective defined units spatial are The,The spatial units are defined by the respective ZIP codes.,300
systems. important in roles plays Stochasticity reaction,Stochasticity plays important roles in reaction systems.,301
the hence are objects. dependencies and neural global applied to determine capture Non-local networks salient,Non-local neural networks are applied to capture global dependencies and hence determine the salient objects.,302
saliency detection algorithm outperforms proposed results state-of-the-art video The experimental methods. the show that,The experimental results show that the proposed algorithm outperforms state-of-the-art video saliency detection methods.,303
(DR) algorithms learning been with accuracy. specialist-level used have retinopathy to detect diabetic Deep,Deep learning algorithms have been used to detect diabetic retinopathy (DR) with specialist-level accuracy.,304
the by reference Grades international panel retinal adjudicated of standard. a served specialists as,Grades adjudicated by a panel of international retinal specialists served as the reference standard.,305
serve DR tool learning valuable as screening. for may Deep algorithms a,Deep learning algorithms may serve as a valuable tool for DR screening.,306
images. very a plays process important remote role of the detection Cloud sensing in,Cloud detection plays a very important role in the process of remote sensing images.,307
"super-pixels Firstly, remote the of into through and SEEDS. sensing segmented images are combination SLIC","Firstly, remote sensing images are segmented into super-pixels through the combination of SLIC and SEEDS.",308
sensing a super-pixel Segmented remote super-pixels compose level database.,Segmented super-pixels compose a super-pixel level remote sensing database.,309
"compared show and results precision achieve recall. better can with conventional methods, HFCNN that, Experimental","Experimental results show that, compared with conventional methods, HFCNN can achieve better precision and recall.",310
burden annotation. alleviates of the manual This,This alleviates the burden of manual annotation.,311
domain trained models examples. performances on test synthetic deteriorates time difference of Such,Such domain difference deteriorates test time performances of models trained on synthetic examples.,312
"Meanwhile, obtain accuracies segmentation on satisfactory fast segmentation models existing to benchmarks. fail usually public","Meanwhile, existing fast segmentation models usually fail to obtain satisfactory segmentation accuracies on public benchmarks.",313
fast networks to of (i.e. learning guide teacher) the segmentation,teacher) to guide the learning of fast segmentation networks (i.e.,314
"other extensive margin. the our experiments, by In large method alternatives proposed a outperforms","In our extensive experiments, the proposed method outperforms other alternatives by a large margin.",315
convolutional neural this for task. widely been have networks (CNNs) used Deep,Deep convolutional neural networks (CNNs) have been widely used for this task.,316
investigate test-time for performance can segmentation. tumor brain We augmentation how improve CNNs',We investigate how test-time augmentation can improve CNNs' performance for brain tumor segmentation.,317
"experiments, an RetinaNet SBL generate SBL-RetinaNet. our detector, to with one-stage trained was In","In our experiments, RetinaNet was trained with SBL to generate an one-stage detector, SBL-RetinaNet.",318
"public image dataset, DOTA. the largest applied SBL-RetinaNet to aerial is existing","SBL-RetinaNet is applied to the largest existing public aerial image dataset, DOTA.",319
plays important in Image role registration images. comparing an,Image registration plays an important role in comparing images.,320
"important images in CT, PET, MRI, particularly etc. is analyzing It medical like","It is particularly important in analyzing medical images like CT, MRI, PET, etc.",321
that normally consistent proposed the asymptotically distributed. is uniformly show We and estimator,We show that the proposed estimator is uniformly consistent and asymptotically normally distributed.,322
"adolescents' illustration, on the risky academic we performance. empirical As investigate of an behaviors impacts","As an empirical illustration, we investigate the impacts of risky behaviors on adolescents' academic performance.",323
steps of distributed different a single into process the model step. This planning unifies the,This model unifies the different steps of the distributed planning process into a single step.,324
no reach goal assume global alone. agent that can the We,We assume that no agent can reach the global goal alone.,325
"optimization been employed problem. resulting Discrete, to optimization the techniques solve successfully have combinatorial,","Discrete, combinatorial, optimization techniques have successfully been employed to solve the resulting optimization problem.",326
used reduction. residual compression for Recent JPEG networks studies deep convolutional have neural artifact (CNNs),Recent studies have used deep residual convolutional neural networks (CNNs) for JPEG compression artifact reduction.,327
proposes This S-Net. CNN a study scalable called,This study proposes a scalable CNN called S-Net.,328
and performance. achieves proposed CNN-based indicate state-of-the-art our approach methods results that outperforms other Experimental,Experimental results indicate that our proposed approach outperforms other CNN-based methods and achieves state-of-the-art performance.,329
systems. the of dynamical to Particle (PF) is used method an estimate often states filtering,Particle filtering (PF) is an often used method to estimate the states of dynamical systems.,330
of the The visualization. method is flow demonstrated quantitative performance for proposed,The performance of the proposed method is demonstrated for quantitative flow visualization.,331
from differentiable are learned are data. modules All and,All modules are differentiable and are learned from data.,332
The system in effectiveness our strategies training the and of are verified experiments.,The effectiveness of our system and training strategies are verified in the experiments.,333
"the are interest scenarios, In many by imaging of noise. corrupted images","In many imaging scenarios, the images of interest are corrupted by noise.",334
"has transform adjustable to distance noise. (SDT), which We, a robustness an stochastic define thus,","We, thus, define a stochastic distance transform (SDT), which has an adjustable robustness to noise.",335
with person synthesize images To is challenging. high-quality poses arbitrary,To synthesize high-quality person images with arbitrary poses is challenging.,336
two two discriminators. generators is adversarial multi-scale a and MsCGAN network consisting of,MsCGAN is a multi-scale adversarial network consisting of two generators and two discriminators.,337
mills. twelve recorded were The datasets from tread last running on subjects,The last twelve datasets were recorded from subjects running on tread mills.,338
of Limitations studied. effects waveform estimation parameter also are communication and channel,Limitations of channel parameter estimation and communication waveform effects are also studied.,339
"Material for understanding functional of analysis modeling, critical is geometric objects. design, and","Material understanding is critical for design, geometric modeling, and analysis of functional objects.",340
ity. and models exhibit variabil- which interesting We focus furniture on structure material,We focus on furniture models which exhibit interesting structure and material variabil- ity.,341
available. dataset will be publicly code The and,The dataset and code will be publicly available.,342
artifacts. methods dual-domain network-based convolutional Several in reducing image neural outstanding compression performance show,Several dual-domain convolutional neural network-based methods show outstanding performance in reducing image compression artifacts.,343
dense the was A performance introduced to extractor DRU. of block in improve,A dense block was introduced to improve the performance of extractor in DRU.,344
of The training. architecture into without integrated U-net is need proposed the module additional,The proposed module is integrated into the U-net architecture without need of additional training.,345
"accuracy the improve of Furthermore, segmentation. spatial to convolution dilated is introduced","Furthermore, dilated convolution is introduced to improve the spatial accuracy of segmentation.",346
We this by is intrinsic find characterized for that preference frequency. spatial system,We find that this system is characterized by intrinsic preference for spatial frequency.,347
for forecasting price evaluation. this and forecasting electricity we structures include In dependency paper,In this paper we include dependency structures for electricity price forecasting and forecasting evaluation.,348
"the hinders Moreover, dimensionality of applications. high SPN large-scale","Moreover, the high dimensionality of SPN hinders large-scale applications.",349
is accurate highly of clusters proposed estimation and The framework in number association. image,The proposed framework is highly accurate in number of clusters estimation and image association.,350
addresses the estimation pixel challenge paper of correspondence between two dense images. This,This paper addresses the challenge of dense pixel correspondence estimation between two images.,351
"data. synthetic unseen, trained to transformations very It good demonstrates performance on realistic, is and","It is trained on synthetic transformations and demonstrates very good performance to unseen, realistic, data.",352
introduce possible belief allowing by over goal flexibility a specification GUSSPs configurations. in goal,GUSSPs introduce flexibility in goal specification by allowing a belief over possible goal configurations.,353
theoretical GUSSP discuss properties. define We and its a formally,We formally define a GUSSP and discuss its theoretical properties.,354
propose approach for of solving determinization a also problems. We class this,We also propose a determinization approach for solving this class of problems.,355
steerable antennas. SUs The equipped with directional are,The SUs are equipped with steerable directional antennas.,356
capacity detection sensing. network formulate of spectrum uses which the for We energy ergodic secondary,We formulate the ergodic capacity of secondary network which uses energy detection for spectrum sensing.,357
"we synthetic trained the using For neural network convolutional localization, images. detection ColorChecker a","For the ColorChecker localization, we trained a detection convolutional neural network using synthetic images.",358
"respect patches with and color all center's the distance. grouped possible to extracted are Subsequently,","Subsequently, all possible color patches are extracted and grouped with respect to the center's distance.",359
"estimation. cost a evaluate the the accuracy Finally, of applied function is to","Finally, a cost function is applied to evaluate the accuracy of the estimation.",360
real and images. synthetic using is method tested The,The method is tested using real and synthetic images.,361
"to invariant proposed overlaps method projections. is and to affine robust The fast,","The proposed method is fast, robust to overlaps and invariant to affine projections.",362
well The case multiple also performs in detection. of ColorCheckers algorithm,The algorithm also performs well in case of multiple ColorCheckers detection.,363
field studies shown can experimental and that Many very rapidly. adaptation have occur,Many experimental and field studies have shown that adaptation can occur very rapidly.,364
how may We also discuss polygenic in of selection detected signals genome. be the the,We also discuss how the signals of polygenic selection may be detected in the genome.,365
"must cooperation, the In co-workers workers behave. how know","In cooperation, the workers must know how co-workers behave.",366
the is important and wild tracking in estimation Multi-person pose and human challenging.,Multi-person human pose estimation and tracking in the wild is important and challenging.,367
"For a are model, training data powerful crucial. training large-scale","For training a powerful model, large-scale training data are crucial.",368
"this network for multi-column inpainting. paper, propose we a In generative image","In this paper, we propose a generative multi-column network for image inpainting.",369
image in synthesizes components network This parallel a within manner stage. one different,This network synthesizes different image components in a parallel manner within one stage.,370
Isoluminant-rainbow colormaps. ensemble the accuracy mean as coloring other to led same,Isoluminant-rainbow coloring led to the same ensemble mean accuracy as other colormaps.,371
"Hue, estimates values. mean of influences ensemble not luminance,","Hue, not luminance, influences ensemble estimates of mean values.",372
model video. accurate helps the An activity objects in moving background detecting of,An accurate background model helps detecting activity of moving objects in the video.,373
types on dynamic problem experiments different conducted related variety of to is A backgrounds.,A variety of experiments is conducted on different problem types related to dynamic backgrounds.,374
"solution. these problems, scratch a from feasible cope To is training detectors with","To cope with these problems, training detectors from scratch is a feasible solution.",375
"In from object we this robustly. to paper, scratch train detectors explore","In this paper, we explore to train object detectors from scratch robustly.",376
segmentation in task is an vision. computer image important Weakly-supervised,Weakly-supervised image segmentation is an important task in computer vision.,377
key quality location is category. from image-level to high A obtain how objects problem,A key problem is how to obtain high quality objects location from image-level category.,378
"guided propose this image paper, we segmentation network a problem. this resolve to In saliency","In this paper, we propose a saliency guided image segmentation network to resolve this problem.",379
We modulator light designed analog a present applications. compressive spatial stochastic imaging for,We present a stochastic analog spatial light modulator designed for compressive imaging applications.,380
this concept demonstrate an in We a imaging using camera. single-pixel system optical,We demonstrate this concept in an optical imaging system using a single-pixel camera.,381
to CNN extract features waterpipes. images contain from was used identified unique to,CNN was used to extract unique features from images identified to contain waterpipes.,382
images with was and distinguish without SVM to built waterpipes. classifier between A,A SVM classifier was built to distinguish between images with and without waterpipes.,383
image-based studies. scope use grow can the to Future research of this method,Future research can use this method to grow the scope of image-based studies.,384
the task validated has volumes. on EM mitochondria The segmenting been in method of,The method has been validated on the task of segmenting mitochondria in EM volumes.,385
can combined method new This matcher. any existing with be minutiae-based,This new method can be combined with any existing minutiae-based matcher.,386
measurements a with range frequencies of the spanning consistent resonance. are over Experimental simulations,Experimental measurements are consistent with simulations over a range of frequencies spanning the resonance.,387
"GA from most wearable on analysis However, focus inertial data methods sensors. existing","However, most existing wearable GA methods focus on data analysis from inertial sensors.",388
"gait is proposed. estimating of parameters spatial-temporal Secondly, a algorithm the for GA","Secondly, a GA algorithm for estimating the spatial-temporal parameters of gait is proposed.",389
image We single shading. decomposing reflectance of a into the address problem and,We address the problem of decomposing a single image into reflectance and shading.,390
smooth local fittings within pixel shading. low-order The of regions for are used pairs,The low-order fittings are used for pixel pairs within local regions of smooth shading.,391
"both they capture shading. global variations of can and structure Together, order the local","Together, they can capture both global order structure and local variations of the shading.",392
"Particularly persists resting associated connectivity, good that a performance into with state. is the sticky","Particularly good performance is associated with a sticky connectivity, that persists into the resting state.",393
on cases in-between is outliers are The wherein normal noises. focus present,The focus is on cases wherein outliers are present in-between normal noises.,394
algorithm resulting The based DPM (DPM-RPF). is termed PF robust,The resulting algorithm is termed DPM based robust PF (DPM-RPF).,395
"signals bases. the modeling, coefficients cannot spanned data LP same by interpolation In distinguish","In data modeling, LP coefficients cannot distinguish signals spanned by the same interpolation bases.",396
relationship establish series. LP also between the Taylor We and,We also establish the relationship between LP and Taylor series.,397
registration global of This the to perform is employed colon positions. representation two first,This representation is employed to first perform global registration of the two colon positions.,398
heritage the all the asset the of Cultural peoples is world. of,Cultural heritage is the asset of all the peoples of the world.,399
cultural inheritance progress is preservation the of and The conducive heritage human of civilization. to,The preservation and inheritance of cultural heritage is conducive to the progress of human civilization.,400
"are creation-era some materials. However, the left reference to due of still lack undetermined with","However, some are still left with creation-era undetermined due to the lack of reference materials.",401
"to be skills. exhibit must In acceptable, order they socio-communicative","In order to be acceptable, they must exhibit socio-communicative skills.",402
is relationship. The geometric typical containment/nesting inter-class a,The containment/nesting is a typical inter-class geometric relationship.,403
and multi-class nested-class-prior loss corresponding function. combined activation the proposing function The its is by,The nested-class-prior is combined by proposing the multi-class activation function and its corresponding loss function.,404
"When on training retrain we the fixed, the model are parameters the dataset. whole","When the parameters are fixed, we retrain the model on the whole training dataset.",405
"testing depend and providing testing limited coverage. fixed a on mainly approaches datasets, pre-defined Existing","Existing testing approaches mainly depend on fixed and pre-defined datasets, providing a limited testing coverage.",406
design with a and present developers. experiences We insights participatory into AI,We present experiences and insights into a participatory design with AI developers.,407
sparsity. these correlation network through part A contains redundancy and significant features of typically,A significant part of these network features typically contains redundancy through correlation and sparsity.,408
"frequent. real the are applications, problems constraints very In monotonicity with","In real applications, the problems with monotonicity constraints are very frequent.",409
useful Changing labels this. set the (relabelling) the of for class data is,Changing the class labels of the data set (relabelling) is useful for this.,410
a under-constrained decomposition image Intrinsic severely problem. is,Intrinsic image decomposition is a severely under-constrained problem.,411
the of can reduce help the interactions decomposition to considerably. ambiguity User,User interactions can help to reduce the ambiguity of the decomposition considerably.,412
method. This of while annotation minimizes maximizing segmentation a the performance image effort human semantic,This minimizes human annotation effort while maximizing the performance of a semantic image segmentation method.,413
on several We our synthetic approach evaluate datasets.,We evaluate our approach on several synthetic datasets.,414
"retrained. neural To tested been an convolutional object, have and networks different classify","To classify an object, different convolutional neural networks have been tested and retrained.",415
with enriched analysis. is concept analysis Soft concept synonymous,Soft concept analysis is synonymous with enriched concept analysis.,416
"calibration Thus, avoid and we handcrafted feature approaches. from error-prone classical hand-eye tracking","Thus, we avoid error-prone hand-eye calibration and handcrafted feature tracking from classical approaches.",417
"WSI a the the approaches, mitotic assess CNN performs segmentation activity. For to both of","For both approaches, the CNN performs a segmentation of the WSI to assess mitotic activity.",418
obtained results show is that the The experimental significantly relevance improved.,The obtained experimental results show that the relevance is significantly improved.,419
R describes the This paper mvLSW. package,This paper describes the R package mvLSW.,420
This accepted the has Statistical Software. Journal paper of by been,This paper has been accepted by the Journal of Statistical Software.,421
present medical We visualization important use feedback expert framework. cases and also for our,We also present important use cases and medical expert feedback for our visualization framework.,422
can assigned transformations images to the misregistration. also to The MvMM embed the correct,The MvMM can also embed transformations assigned to the images to correct the misregistration.,423
for iterated is algorithm The optimization. mode conditional adopted,The iterated conditional mode algorithm is adopted for optimization.,424
"quantification the LA projects segmentation onto of analysis It the and surface, for resulting scarring.","It projects the resulting segmentation onto the LA surface, for quantification and analysis of scarring.",425
achieved in progress Planning years. significant recent has,Planning has achieved significant progress in recent years.,426
novel parameterization. work framework for mesh This spherical presents a,This work presents a novel framework for spherical mesh parameterization.,427
histology a breast cancer. analysis Microscopic image detection in early of cornerstone is,Microscopic histology image analysis is a cornerstone in early detection of breast cancer.,428
Thus high this is in process demand. automating,Thus automating this process is in high demand.,429
"design cooperative a a realistic Quizbowl. We question task, using human-computer answering grounded, setting","We design a grounded, realistic human-computer cooperative setting using a question answering task, Quizbowl.",430
also guidance human-in-the-loop language provide design We natural settings. for processing,We also provide design guidance for natural language processing human-in-the-loop settings.,431
Models long-time the numerically behaviour to analysed examine and solved and are distribution. auxin,Models are analysed and solved numerically to examine the long-time behaviour and auxin distribution.,432
region feasibility recognition. using the paper This expression for investigates of the periocular,This paper investigates the feasibility of using the periocular region for expression recognition.,433
solve Most face. by works this tried the to have whole analyzing,Most works have tried to solve this by analyzing the whole face.,434
of the the immediate facial eye. vicinity is in the Periocular region,Periocular is the facial region in the immediate vicinity of the eye.,435
introducing acquisition network achieve We design. data goal by and training innovations on this several,We achieve this goal by introducing several innovations on training data acquisition and network design.,436
styles. classifying artistic mostly into been on different art artworks has focused analysis Automatic,Automatic art analysis has been mostly focused on classifying artworks into different artistic styles.,437
"We for present multi-modal art semantic understanding. dataset SemArt, a","We present SemArt, a multi-modal dataset for semantic art understanding.",438
"understanding when art against of compared show models human our remarkable levels Moreover, evaluation.","Moreover, our models show remarkable levels of art understanding when compared against human evaluation.",439
simple is shallow. proposed The and network,The proposed network is simple and shallow.,440
"deep image-level propose a method visualization bioresorbable learning-based we only scaffold for labels. Therefore, using","Therefore, we propose a deep learning-based method for bioresorbable scaffold visualization using only image-level labels.",441
one is in sciences. Microscopy most life of essential imaging techniques the,Microscopy is one of the most essential imaging techniques in life sciences.,442
in to research solve High-quality (potentially life-saving) images are problems. biomedical order required,High-quality images are required in order to solve (potentially life-saving) biomedical research problems.,443
reduce or effectiveness. model in their PSF will estimated the errors Any noise,Any errors in the estimated PSF or noise model will reduce their effectiveness.,444
We formulate programming NP-hardness. the binary and non-linear problem prove BP as a its problem,We formulate the BP problem as a non-linear binary programming problem and prove its NP-hardness.,445
as applied illustrative voltage inverter a source is an example. This droop-controlled to theory,This theory is applied to a droop-controlled voltage source inverter as an illustrative example.,446
a information work presents fusion This for source method localization applications. in,This work presents a method for information fusion in source localization applications.,447
computed tomography identification decomposition. potential has material great Spectral in (CT) a and,Spectral computed tomography (CT) has a great potential in material identification and decomposition.,448
"reconstruction reconstruction Then, named basic technique algebraic a material (MSART). method simultaneous develop we material","Then, we develop a basic material reconstruction method named material simultaneous algebraic reconstruction technique (MSART).",449
"subtraction. we present background a this the In letter, approach in implementing novel","In this letter, we present a novel approach in implementing the background subtraction.",450
"for grayscale The method compressed, image the subtraction. background low-resolution uses proposed","The proposed method uses compressed, low-resolution grayscale image for the background subtraction.",451
salient images well. found very low-resolution the to These grayscale preserve information were,These low-resolution grayscale images were found to preserve the salient information very well.,452
The the of results proposed methodology the confirm effectiveness approach. our of,The results of the proposed methodology confirm the effectiveness of our approach.,453
The by will such of information centralized tags a be server. maintained all,The information of all such tags will be maintained by a centralized server.,454
security would necessary activated system found is car. be for the Once arresting it,Once it is found necessary security system would be activated for arresting the car.,455
from using networks However parameters. usually by too suffer overfitting much the,However the networks usually suffer from overfitting by using too much parameters.,456
use deal Dropout method overfitting. is to with widely a,Dropout is a widely use method to deal with overfitting.,457
"track this To DropFilter, for dropout convolutional layers. method we propose a problem, new","To track this problem, we propose DropFilter, a new dropout method for convolutional layers.",458
DropFilter filters. the outputs of some suppresses randomly,DropFilter randomly suppresses the outputs of some filters.,459
"convolutional of DropFilter, remarkably and networks performance we Using ImageNet. the improve CIFAR on","Using DropFilter, we remarkably improve the performance of convolutional networks on CIFAR and ImageNet.",460
sets; was competitive approach was it sequestered third. ranked on Our,Our approach was competitive on sequestered sets; it was ranked third.,461
our the Experiments of on VOC SeeNet methods. state-of-the-art other well Pascal over superiority demonstrate,Experiments on Pascal VOC well demonstrate the superiority of our SeeNet over other state-of-the-art methods.,462
network have of Principles been the topology widely connectome. human studied in,Principles of network topology have been widely studied in the human connectome.,463
"real Moreover, outperforms data synthetic KITTI with data combined SDR real alone. data KITTI","Moreover, synthetic SDR data combined with real KITTI data outperforms real KITTI data alone.",464
recently. has interests segmentation vision the one in computer been of research Semantic leading,Semantic segmentation has been one of the leading research interests in computer vision recently.,465
"and for driving. foundation such as a autonomous robotics as serves It perception many fields,","It serves as a perception foundation for many fields, such as robotics and autonomous driving.",466
bulk power significantly disrupt delivery. of power transformer a failure can sudden A,A sudden failure of a power transformer can significantly disrupt bulk power delivery.,467
"band small. normal the is the SNR width operations, During of","During normal operations, the width of the SNR band is small.",468
"present approach, different a the capabilities. We by aforementioned inspired human","We present a different approach, inspired by the aforementioned human capabilities.",469
the question. approach on based is structure the The compositional of,The approach is based on the compositional structure of the question.,470
of answered corresponding to its substructures. by be procedures can a composition consequently The question,The question can consequently be answered by a composition of procedures corresponding to its substructures.,471
"representation object a classes, patterns for include and These properties relations. parametric","These patterns include a parametric representation for object classes, properties and relations.",472
"base an external In for queried is addition, common-knowledge. required knowledge","In addition, an external knowledge base is queried for required common-knowledge.",473
resonance the disease. in and management magnetic Cardiac diagnosis (CMR) is of extensively cardiovascular used,Cardiac magnetic resonance (CMR) is used extensively in the diagnosis and management of cardiovascular disease.,474
"this methods to existing pay However, resolution attention discrepancy person ReID problem. most little","However, most existing person ReID methods pay little attention to this resolution discrepancy problem.",475
original according images to and resolutions. their first training categorize quantify We,We first quantify and categorize original training images according to their resolutions.,476
"previous and state-of-the-art methods a large our whistles, by outperforms margin. Without approach bells","Without bells and whistles, our approach outperforms previous state-of-the-art methods by a large margin.",477
requires This memory similarity compact computation. efficient representation storage minimal enables and,This compact representation requires minimal memory storage and enables efficient similarity computation.,478
"vision drones, as smartphones, headsets such and Mobile our are revolutionizing systems augmented-reality lives.","Mobile vision systems such as smartphones, drones, and augmented-reality headsets are revolutionizing our lives.",479
trade-offs. offer each resource-accuracy NestDNN deep to learning flexible enables model,NestDNN enables each deep learning model to offer flexible resource-accuracy trade-offs.,480
manners. Fault analysis often are bad in processed and data separate,Fault analysis and bad data are often processed in separate manners.,481
The the based is design in theory observer paper. approach presented this on,The design approach based on the observer theory is presented in this paper.,482
in eigenvalues the excessive subspace. it designing in However the freedom loses,However it loses freedom in designing the eigenvalues in the excessive subspace.,483
"general, While to design theoretical analysis dedicated the lines. transmission framework and are is the","While the theoretical framework is general, the analysis and design are dedicated to transmission lines.",484
"exchange so, By information overheads. it performance strikes and good tradeoff between gains a doing","By doing so, it strikes a good tradeoff between performance gains and information exchange overheads.",485
in role analysis segmentation Cloud modeling. crucial climate for a image plays,Cloud segmentation plays a crucial role in image analysis for climate modeling.,486
training for segmentation and labeling Manually the data time-consuming is cloud error-prone.,Manually labeling the training data for cloud segmentation is time-consuming and error-prone.,487
extensively been analyze have to and Adversarial synthesize used biomedical images. techniques,Adversarial techniques have been extensively used to synthesize and analyze biomedical images.,488
broadcasting and cameras sports Calibrating important for analysis. sports autonomous is,Calibrating sports cameras is important for autonomous broadcasting and sports analysis.,489
"a First, pose engine. we camera develop novel","First, we develop a novel camera pose engine.",490
synthetic real both our on We and evaluate method data.,We evaluate our method on both synthetic and real data.,491
in used Supervised are methods learning. widely machine learning,Supervised learning methods are widely used in machine learning.,492
"limits in data of lack these However, the technologies. existing application the labels of","However, the lack of labels in existing data limits the application of these technologies.",493
"of visual cognition that, and VIL. perception two main processes are Note","Note that, perception and cognition are two main visual processes of VIL.",494
the Analyzing process. fashion essential in design is fashion attributes,Analyzing fashion attributes is essential in the fashion design process.,495
"different ability such is careers In fashion in useful abstraction with fact, purposes. many","In fact, such abstraction ability is useful in many fashion careers with different purposes.",496
"fashion a this In recognizing propose project, for we approach data-driven attributes.","In this project, we propose a data-driven approach for recognizing fashion attributes.",497
"becomes the large channel In non-stationary. spatially MIMO a with system arrays, massive","In a massive MIMO system with large arrays, the channel becomes spatially non-stationary.",498
equivalent also precoders. SINR approximate ZF provide of of deterministic We the an,We also provide an approximate deterministic equivalent of the SINR of ZF precoders.,499
is screening. applied of breast tools one cancer early the Mammography most commonly for,Mammography is one of the most commonly applied tools for early breast cancer screening.,500
patches with by automatically. or methods image mainly these deal extracting challenges mass-centered manually Existing,Existing methods deal with these challenges mainly by extracting mass-centered image patches manually or automatically.,501
"have to block designed Especially, three AU merits. is the","Especially, the AU block is designed to have three merits.",502
"information loss by compensates of Firstly, the upsampling bilinear it upsampling. dense","Firstly, it compensates the information loss of bilinear upsampling by dense upsampling.",503
"method Secondly, effective features. designs and a more low-level high- it fuse to","Secondly, it designs a more effective method to fuse high- and low-level features.",504
"highlight function Thirdly, a rich-information includes channels. it channel-attention to","Thirdly, it includes a channel-attention function to highlight rich-information channels.",505
"publicly We CBIS-DDSM available evaluated the on proposed two INbreast. method datasets, and","We evaluated the proposed method on two publicly available datasets, CBIS-DDSM and INbreast.",506
"a NDA is this paper, for In proposed. algorithm QPSK SNR signal estimation","In this paper, a NDA SNR estimation algorithm for QPSK signal is proposed.",507
that vision is technology. for failures to are This inherent detecting necessary,This is necessary for detecting failures that are inherent to vision technology.,508
unsurpassed MIMO bound. equivalent The was (WF-MIMO) as channel water-filling SE capacity an upper treated,The equivalent MIMO water-filling (WF-MIMO) channel capacity was treated as an unsurpassed SE upper bound.,509
"benchmark Recently, proposed performance researchers schemes trying the approach the to bound. various have and","Recently, researchers have proposed various schemes trying to approach the benchmark and the performance bound.",510
"are (RF) the with mmWave of they radio-frequency But, limit chains? MIMO real systems reduced","But, are they the real limit of mmWave MIMO systems with reduced radio-frequency (RF) chains?",511
"That bound GBMM the is, SE. renews the upper of","That is, GBMM renews the upper bound of the SE.",512
derived SNR The the in regime. best solution is high,The best solution is derived in the high SNR regime.,513
We realize optimized coding precoder the activation. approach propose a to,We propose a coding approach to realize the optimized precoder activation.,514
extension is also An to communications discussed. broadband mmWave,An extension to mmWave broadband communications is also discussed.,515
"date, consecutive flow estimation optical into pairs frames take only methods account. top-performing of To","To date, top-performing optical flow estimation methods only take pairs of consecutive frames into account.",516
"in conducted by Safety BSMs Pilot the were used. generated Michigan, Arbor, Model Deployment Ann","BSMs generated by the Safety Pilot Model Deployment conducted in Ann Arbor, Michigan, were used.",517
module train specialized multiple propose networks to instead. We,We propose to train multiple specialized module networks instead.,518
the is subset module classes. trained of to recognize a Each action,Each module is trained to recognize a subset of the action classes.,519
data-driven techniques. method geometric that a We traditional and understanding introduce couples,We introduce a method that couples traditional geometric understanding and data-driven techniques.,520
"antenna be as In addition, spatial can and transmission per full achieved well. diversity equal-power","In addition, equal-power transmission per antenna and full spatial diversity can be achieved as well.",521
problem using the student-project We describe simulated a allocation annealing. to solution,We describe a solution to the student-project allocation problem using simulated annealing.,522
"has an iterative algorithm the the to If axis axis. developed unknown, estimate is been","If the axis is unknown, an iterative algorithm has been developed to estimate the axis.",523
simulated The using alpha-helices. is methodology illustrated protein,The methodology is illustrated using simulated protein alpha-helices.,524
consider the of problem person re-identification. We video-based,We consider the problem of video-based person re-identification.,525
a videos is person goal from captured cameras. to different under identify The,The goal is to identify a person from videos captured under different cameras.,526
frame on generates each based features. score method frame-level attention an Our for,Our method generates an attention score for each frame based on frame-level features.,527
performance. Extensive the achieves experiments datasets state-of-the-art demonstrate on benchmark method our that two,Extensive experiments on two benchmark datasets demonstrate that our method achieves the state-of-the-art performance.,528
goal if exposure present The know the distortions. is to video,The goal is to know if the video present exposure distortions.,529
"system. important local a for as kind human visual provide information variations, Image of textures,","Image textures, as a kind of local variations, provide important information for human visual system.",530
technique develop We the threshold accuracy. to of a mechanism improve adaptive recognition,We develop a technique of adaptive threshold mechanism to improve the recognition accuracy.,531
paper this the derived proofs functions. are in the Dempster-Shafer of framework All in belief,All the proofs in this paper are derived in the framework of Dempster-Shafer belief functions.,532
"solution iterative through Source Achieves Model. a the Graphical Hermitian Sparse Super-Resolution, Second: of","Second: Achieves Super-Resolution, through the iterative solution of a Sparse Hermitian Source Graphical Model.",533
Dimensional High Complex the efficiently in of Tackles Third: estimation set up connectivity. and,Third: Tackles efficiently in High Dimensional and Complex set up the estimation of connectivity.,534
Association. two employ Object methods Detection steps: and MOT Data Most,Most MOT methods employ two steps: Object Detection and Data Association.,535
few the years deep in detection learning. tremendous last progress made Object has to due,Object detection has made tremendous progress in the last few years due to deep learning.,536
the to affinities between objects frames. different in compute,to compute affinities between the objects in different frames.,537
appearing frames. also multiple and between for video DAN disappearing accounts objects,DAN also accounts for multiple objects appearing and disappearing between video frames.,538
"of the and of formulation our ranging. simulations Finally, we lidar demonstrate via effectiveness algorithm","Finally, we demonstrate the effectiveness of our formulation and algorithm via simulations of lidar ranging.",539
"bootstrap illustrate studies. Moreover, methods develop we via numerical inference and our","Moreover, we develop bootstrap inference and illustrate our methods via numerical studies.",540
through motions signals. WiFi be recognized human that and positions Recent research can shown has,Recent research has shown that human motions and positions can be recognized through WiFi signals.,541
"In propose paper, we human this system. activities recognition WiFi-based a Wi-Motion,","In this paper, we propose Wi-Motion, a WiFi-based human activities recognition system.",542
Definite (SPD) space The Symmetric matrices matrices. geometry is that the covariance of Positive of,The space geometry of the covariance matrices is that of Symmetric Positive Definite (SPD) matrices.,543
The a single multiple processes independent problem anomalous considered. of is detecting among process,The problem of detecting a single anomalous process among multiple independent processes is considered.,544
demonstrate performance for finite Simulation strong results both regime scenarios. in the,Simulation results demonstrate strong performance in the finite regime for both scenarios.,545
"and necessary image The data diagnosis, lesion is within prognosis. for of identification treatment medical","The identification of lesion within medical image data is necessary for diagnosis, treatment and prognosis.",546
"the knowledge. specialized laborious images medical highly is labeling requiring in lesion However,","However, labeling the lesion in medical images is laborious requiring highly specialized knowledge.",547
"frames. could and also contain lot video of sequence irrelevant a a redundant Nevertheless,","Nevertheless, a video sequence could also contain a lot of redundant and irrelevant frames.",548
to of a interest task. challenging How critical is information locate,How to locate critical information of interest is a challenging task.,549
maps out modulating threshing and Suppression the by achieved activations. feature is weak,Suppression is achieved by modulating the feature maps and threshing out weak activations.,550
next updated feature then The maps used in iteration. are the,The updated feature maps are then used in the next iteration.,551
"multiple video is on summaries. based Finally, classified the","Finally, the video is classified based on multiple summaries.",552
classification datasets. method achieves video multiple out- in standing proposed The performances,The proposed method achieves out- standing performances in multiple video classification datasets.,553
"addition, nearly removes our all method In tuning. hyperparameters","In addition, our method nearly removes all hyperparameters tuning.",554
methods Analysis Decision and fields widely used in Multi-Criteria (MCDA) various disciplines. are,Multi-Criteria Decision Analysis (MCDA) methods are widely used in various fields and disciplines.,555
"as MCDA methods of different recommendations, results. improper inconsistent application Their deliver the decreases quality","Their improper application decreases the quality of recommendations, as different MCDA methods deliver inconsistent results.",556
at framework for web public a platform proposed was use implemented The within available www.mcda.it.,The proposed framework was implemented within a web platform available for public use at www.mcda.it.,557
to more learning. One they transfer of amenable that policies is advantage are deep reactive,One advantage of deep reactive policies is that they are more amenable to transfer learning.,558
on underscore of our three benchmark the Experiments domains algorithm. transfer value different,Experiments on three different benchmark domains underscore the value of our transfer algorithm.,559
"unavailable. collected, dataset independently for are were and HS-SWIR the The datasets metadata","The datasets were independently collected, and metadata for the HS-SWIR dataset are unavailable.",560
local often on overlook are built (e.g. models the and challenges Current generative key convolutions,Current generative models are often built on local convolutions and overlook the key challenges (e.g.,561
of given degrees different Warping-GAN transformation is distinct controlling poses. capable target,Warping-GAN is capable of controlling different transformation degrees given distinct target poses.,562
"any Moreover, enough light-weight proposed into the to is networks. be flexible warping-block and injected","Moreover, the proposed warping-block is light-weight and flexible enough to be injected into any networks.",563
this enough attenuation. without DNN retraining several single to any We noise concatenate provide times,We concatenate this single DNN several times without any retraining to provide enough noise attenuation.,564
planning (MTS) inspiration This from Search Target algorithm Moving algorithms. draws,This planning algorithm draws inspiration from Moving Target Search (MTS) algorithms.,565
"results effectiveness Finally, evaluation we the show of that demonstrate our approach.","Finally, we show evaluation results that demonstrate the effectiveness of our approach.",566
quality images data. problem from the address We high MRI reconstructing of undersampled,We address the problem of reconstructing high quality images from undersampled MRI data.,567
problem. highly This due task to nature challenging a the of ill-posed the is,This is a challenging task due to the highly ill-posed nature of the problem.,568
"AI of one (VA) in priorities alignment"" research. as is ""Value top the considered","""Value alignment"" (VA) is considered as one of the top priorities in AI research.",569
"""mimetic"" forms begins distinct by isolating and of paper VA: two ""anchored."" The","The paper begins by isolating two distinct forms of VA: ""mimetic"" and ""anchored.""",570
approach better which it Then naturalistic VA discusses fallacy. the avoids,Then it discusses which VA approach better avoids the naturalistic fallacy.,571
the for approaches implications VA The discussion that stumbling blocks reveals of neglect naturalistic fallacy.,The discussion reveals stumbling blocks for VA approaches that neglect implications of the naturalistic fallacy.,572
this instantiation attracted from fourth academia the industry. both workshop interest and of The significant,The fourth instantiation of this workshop attracted significant interest from both academia and the industry.,573
"an propose approach. To analysis address we instant efficient critical this,","To address this, we propose an efficient critical instant analysis approach.",574
demonstrate a mechanical method our of and it validate models variety on We tests. with,We demonstrate our method on a variety of models and validate it with mechanical tests.,575
model predominant approaches for compression. one Channel is of pruning the deep,Channel pruning is one of the predominant approaches for deep model compression.,576
the method. effectiveness of Extensive experiments demonstrate our,Extensive experiments demonstrate the effectiveness of our method.,577
"two out forehead key point nose-mouth. detection Initially, carried is facial subregions: and on","Initially, key point detection is carried out on two facial subregions: forehead and nose-mouth.",578
"interest. unstable removing the of After to filtering features, is temporal applied frequencies isolate","After removing unstable features, the temporal filtering is applied to isolate frequencies of interest.",579
This the component. enable each peaks locate in positions to of will,This will enable to locate the positions of peaks in each component.,580
"resulting II the Type-I, we and design problem, Through three filters: prototype III.","Through the resulting problem, we design three prototype filters: Type-I, II and III.",581
with gaps. presented stimulus Most a from gathering decisions require different information,Most decisions require information gathering from a stimulus presented with different gaps.,582
"ambiguous. process Indeed, rarely integration the brain this of is","Indeed, the brain process of this integration is rarely ambiguous.",583
"the TanDEM-X with SAR map terrain unprecedented accuracy. Recently, provided interferometry a global mission bistatic","Recently, the bistatic SAR interferometry mission TanDEM-X provided a global terrain map with unprecedented accuracy.",584
This of is Weighted previously usually done DEM (WA) cells. by aligned Averaging,This is usually done by Weighted Averaging (WA) of previously aligned DEM cells.,585
a fully is The Neural Network (ANN). modeled connected using Artificial prediction,The prediction is modeled using a fully connected Artificial Neural Network (ANN).,586
learning re-visits the problem by approach. an binarization introducing The adversarial paper,The paper re-visits the binarization problem by introducing an adversarial learning approach.,587
the networks adversarial increase By we can the jointly two our of system. training robustness,By jointly training the two networks we can increase the adversarial robustness of our system.,588
"learn noteworthy it model from that Also, is our unpaired can data.","Also, it is noteworthy that our model can learn from unpaired data.",589
method suggest superior the Experimental proposed datasets. used that achieves performance results over DIBCO widely,Experimental results suggest that the proposed method achieves superior performance over widely used DIBCO datasets.,590
long-range is Learning fundamental relations image/video to recognition. to capture,Learning to capture long-range relations is fundamental to image/video recognition.,591
"observation remote volume huge diverse data. provide Currently, numerous satellites sensing a of earth","Currently, numerous remote sensing satellites provide a huge volume of diverse earth observation data.",592
"of is this For semi-global multi-sensor in the applicability this matching investigated unconventional setting. purpose,","For this purpose, the applicability of semi-global matching is investigated in this unconventional multi-sensor setting.",593
"the illumination sensitive to is However, changes. noise and LBP","However, the LBP is sensitive to noise and illumination changes.",594
zero a We trigonometric the level set curve model as of the polynomial.,We model the curve as the zero level set of a trigonometric polynomial.,595
images. RadSynth synthetic traditional entropy images produced excellent compared GLCM to entropy,RadSynth produced excellent synthetic entropy images compared to traditional GLCM entropy images.,596
"In a single can pixel. a one observed in setting, be general surface more than","In a general setting, more than one surface can be observed in a single pixel.",597
stitching by image limitations these We optimization. in boundaries with regular unified formulating a address,We address these limitations by formulating image stitching with regular boundaries in a unified optimization.,598
"to Moreover, among lead cost will motion motion patterns recognition. users irregular for computation expensive","Moreover, irregular motion patterns among users will lead to expensive computation cost for motion recognition.",599
"monotone accelerating algorithms package for iterative discuss SQUAREM slow, R which We exhibit convergence.","We discuss R package SQUAREM for accelerating iterative algorithms which exhibit slow, monotone convergence.",600
EM-like provides algorithms. of iterative algorithm speed-up Squared scheme (Squarem) significant,Squared iterative scheme (Squarem) algorithm provides significant speed-up of EM-like algorithms.,601
MRFs to algorithms. in general in intractable work approximation resorts related and Inference is,Inference in MRFs is intractable in general and related work resorts to approximation algorithms.,602
a and second Our to is contribution faster develop approach. second,Our second contribution is to develop a second and faster approach.,603
"sub-optimality leverage provide Moreover, from to results we recent optimization guarantees. theory per-instance","Moreover, we leverage recent results from optimization theory to provide per-instance sub-optimality guarantees.",604
multi-class in segmentation the image We proposed approaches problems. demonstrate,We demonstrate the proposed approaches in multi-class image segmentation problems.,605
of spontaneous Epilepsy increase persistent the creates probability in seizures. a,Epilepsy creates a persistent increase in the probability of spontaneous seizures.,606
videos. effective paper proposes for surveillance analysis a novel This security event in framework,This paper proposes a novel effective framework for security event analysis in surveillance videos.,607
"as objects recognized the owners Second, monitored and well. in the real-time of are","Second, the owners of the objects are recognized and monitored in real-time as well.",608
"will he/she this is person moves object, verified owner. any whether anyone If its be","If anyone moves any object, this person will be verified whether he/she is its owner.",609
We impact artefacts results. controls on of the reconstruction segmentation discuss achievable further the and,We further discuss the impact of reconstruction artefacts and segmentation controls on the achievable results.,610
"techniques. for quality an there urgent automatic image is assessment Therefore, need","Therefore, there is an urgent need for automatic image quality assessment techniques.",611
a a on artefact synthetic based severity. predetermined also using learning approach curriculum We investigate,We also investigate a learning approach using a predetermined curriculum based on synthetic artefact severity.,612
of methods. approach to a We quality assessment state-of-the-art our compare range,We compare our approach to a range of state-of-the-art quality assessment methods.,613
three proposed results that clearly the achieves method state-of-the-art the on Experimental results datasets. demonstrate,Experimental results clearly demonstrate that the proposed method achieves the state-of-the-art results on three datasets.,614
tracking is Target videos topic. new in research a hyperspectral,Target tracking in hyperspectral videos is a new research topic.,615
"neighboring is targets step, adopted the to tracking KCF from environment. distinguish In","In the tracking step, KCF is adopted to distinguish targets from neighboring environment.",616
"of been an there Over recent on has years, vehicles. studies autonomous the explosion","Over the recent years, there has been an explosion of studies on autonomous vehicles.",617
from Many of amount human data drivers. large collected,Many collected large amount of data from human drivers.,618
train networks. ways or supervised neural unsupervised alternative This to provides deep,This provides alternative supervised or unsupervised ways to train deep neural networks.,619
potential and Its needs. accelerate collide reinforcement prepares deep the system for to naturalness learning,Its naturalness to accelerate and collide prepares the system for potential deep reinforcement learning needs.,620
pose important Camera in vision. estimation is problem an computer,Camera pose estimation is an important problem in computer vision.,621
"establish a alternative years, such popular correspondences. forests regression recent have become In to","In recent years, regression forests have become a popular alternative to establish such correspondences.",622
studies (MIMO) receive system. the massive multiple-output This selection paper multiple-input antenna in,This paper studies the receive antenna selection in massive multiple-input multiple-output (MIMO) system.,623
performance challenge of face Deeper (ii) the networks saturation.,(ii) Deeper networks face the challenge of performance saturation.,624
the for denoising. ECNDNet outperforms demonstrate the image experiments state-of-the-art Extensive methods that,Extensive experiments demonstrate that the ECNDNet outperforms the state-of-the-art methods for image denoising.,625
prevalence that in results This in fluctuations facilitate large extinction.,This results in large fluctuations in prevalence that facilitate extinction.,626
"problem. learning is zero-shot way, supervised learning this into turned a In","In this way, zero-shot learning is turned into a supervised learning problem.",627
method. benchmark show datasets four superiority the Experiments proposed our of on,Experiments on four benchmark datasets show the superiority of our proposed method.,628
"To this, world's for open-ended the created largest TallyQA, we dataset do counting.","To do this, we created TallyQA, the world's largest dataset for open-ended counting.",629
proposals. that new uses region algorithm We with relation a counting for propose networks,We propose a new algorithm for counting that uses relation networks with region proposals.,630
lets efficiently Our high-resolution imagery. with method used relation be networks,Our method lets relation networks be efficiently used with high-resolution imagery.,631
process and mutual nonverbal emotional Human in dialogs a adaptation. influence dyadic is of communication,Human nonverbal emotional communication in dyadic dialogs is a process of mutual influence and adaptation.,632
"clearly emotions distinct be might First, visible. not","First, distinct emotions might not be clearly visible.",633
"cause-effect variant time. transient over participants is relation Second, and","Second, participants cause-effect relation is transient and variant over time.",634
semantically between for present similar networks obtaining dense recurrent correspondences We (RTNs) images. transformer,We present recurrent transformer networks (RTNs) for obtaining dense correspondences between semantically similar images.,635
"several state-of-the-art semantic attained correspondence. RTNs, With is on benchmarks for performance","With RTNs, state-of-the-art performance is attained on several benchmarks for semantic correspondence.",636
"reconstruction Nevertheless, problems. to model very has limited application seen tomographic this","Nevertheless, this model has seen very limited application to tomographic reconstruction problems.",637
This using paper description. manipulating language natural addresses the problem of images,This paper addresses the problem of manipulating images using natural language description.,638
sampling calibration the take to focus both step the The and scheme beam time.,The calibration step to focus the beam and the sampling scheme both take time.,639
"preliminary method this on theory. scanning a based we In work, compressive sampling propose","In this preliminary work, we propose a scanning method based on compressive sampling theory.",640
We new with incremental algorithm deblurring a image automatic for aggregation multi-image propose selection.,We propose a new incremental aggregation algorithm for multi-image deblurring with automatic image selection.,641
"steps multi-image a we Here, deblurring the two process. as problem approach","Here, we approach the multi-image deblurring problem as a two steps process.",642
datasets. We on findings our validate real several synthetic and,We validate our findings on several synthetic and real datasets.,643
on recreational increasingly are used Vehicles Aerial Micro Cameras for mounted (MAVs) photography.,Cameras mounted on Micro Aerial Vehicles (MAVs) are increasingly used for recreational photography.,644
"discourage estimation. locally changes Moreover, attacks filter use proposed parameters the parameter to its that","Moreover, the proposed filter locally changes its parameters to discourage attacks that use parameter estimation.",645
classic sets. are features demonstrated novel data to The the of package through and application,The features of the package are demonstrated through application to classic and novel data sets.,646
"approach for this work, lightweight in propose better CNNs. context we In a exploitation simple,","In this work, we propose a simple, lightweight approach for better context exploitation in CNNs.",647
"this replacement causal However, debated. mechanism to be the continues for","However, the causal mechanism for this replacement continues to be debated.",648
present results and genetic comparison archaeological These for data. opportunities with useful,These results present useful opportunities for comparison with archaeological and genetic data.,649
"drawback. is to this In MPPT novel two-stage this paper, a presented method overcome","In this paper, a novel two-stage MPPT method is presented to overcome this drawback.",650
is then P&O re-activated trace algorithm changes MPP. to small of the new,P&O algorithm is then re-activated to trace small changes of the new MPP.,651
and the its simple. Open method operation makes implementation loop proposed cheap of,Open loop operation of the proposed method makes its implementation cheap and simple.,652
results validate experimental methods. Simulations proposed the in the performance Matlab/Simulink and of,Simulations in Matlab/Simulink and experimental results validate the performance of the proposed methods.,653
"generative two dimensions latent with per probability uni-modal class, density. The net only is using","The net is using only two latent generative dimensions per class, with uni-modal probability density.",654
"Only be training are which images well perturbed. reconstructed, can","Only training images which are well reconstructed, can be perturbed.",655
"perturbations, the For drifts empirical tentative images. of we reconstructed good well use","For good perturbations, we use the tentative empirical drifts of well reconstructed images.",656
The predictive is sense. statistical in not usual the construct,The construct is not predictive in the usual statistical sense.,657
of convolutional basic learn DNN defects. features parts to is The fully presented,The fully convolutional DNN is presented to learn basic features of parts defects.,658
network shallow empirical these as thresholds. and with weights assemble We a methods fixed,We assemble these methods as a shallow network with fixed weights and empirical thresholds.,659
training. thresholds enhance ability then and its These realize to end-to-end released adaptation are,These thresholds are then released to enhance its adaptation ability and realize end-to-end training.,660
We testing. demonstrate experiment also samples for an only with intermediates,We also demonstrate an experiment with only intermediates samples for testing.,661
"non-mutually ""treatments"". and/or exclusive (combinations of) continuous, Here may discrete, X include","Here X may include (combinations of) continuous, discrete, and/or non-mutually exclusive ""treatments"".",662
treatment vector procedure of) X the our average (multi-valued) When effect(s). (a binary-valued is recovers,When X is binary-valued (multi-valued) our procedure recovers the (a vector of) average treatment effect(s).,663
"compared to manner. reported, Significant fully-supervised networks performance in state-of-art is trained improvement a segmentation","Significant performance improvement is reported, compared to state-of-art segmentation networks trained in a fully-supervised manner.",664
"with entities financial may resources. they by be In large fact, aided and human","In fact, they may be aided by entities with large financial and human resources.",665
states are states less-secure our deemed These `risky' as in approach.,These less-secure states are deemed as `risky' states in our approach.,666
show in results `risky' future identifying well performs Experimental that states. the proposed framework,Experimental results show that the proposed framework performs well in identifying future `risky' states.,667
dataset practice. to functionality the NHANES muvis exploited declare of the We in,We exploited the NHANES dataset to declare the functionality of muvis in practice.,668
"detection In paper, bacterial an this endomicroscopy approach images. optical for we unsupervised in propose","In this paper, we propose an unsupervised approach for bacterial detection in optical endomicroscopy images.",669
"In the S-SPADE. propose version paper, of we a present new","In the present paper, we propose a new version of S-SPADE.",670
paper This novel to solve a proposed predictive processing these method issues. signal,This paper proposed a novel predictive signal processing method to solve these issues.,671
"In significant led image in sparsity-driven past decade, the reconstruction. to has improvements regularization","In the past decade, sparsity-driven regularization has led to significant improvements in image reconstruction.",672
"such total rely as of sparsity. (TV), on variation regularizers, analytical models Traditional","Traditional regularizers, such as total variation (TV), rely on analytical models of sparsity.",673
"inspired from deep is towards moving models, trainable the However, learning. increasingly field","However, increasingly the field is moving towards trainable models, inspired from deep learning.",674
DIP regularization. combining with paper This framework extends TV traditional by the the it,This paper extends the DIP framework by combining it with the traditional TV regularization.,675
ask we from to we alone? signals brain see what recognize computers Can,Can we ask computers to recognize what we see from brain signals alone?,676
"significantly state-of-the-art obtains models. existing surpassing proposed results, The novel architecture other","The proposed novel architecture obtains state-of-the-art results, significantly surpassing other existing models.",677
of the glaucoma feasibility Conclusion: in grading experiments deep technology the demonstrated severity. The learning,Conclusion: The experiments demonstrated the feasibility of the deep learning technology in grading glaucoma severity.,678
"fail brain neuroactivity. analysis capture linear techniques may However, of nonlinearity to dimensionality","However, linear dimensionality analysis techniques may fail to capture nonlinearity of brain neuroactivity.",679
networks. to is convolutional needed regularize dropout structured a of Thus form,Thus a structured form of dropout is needed to regularize convolutional networks.,680
networks. Extensive works that show than in better dropout convolutional DropBlock experiments regularizing,Extensive experiments show that DropBlock works better than dropout in regularizing convolutional networks.,681
"experiments, in found cells been organize such In aggregates. have many to","In many such experiments, cells have been found to organize in aggregates.",682
cancer breast present in cultures vitro exhibiting cells We patterns. of of new three-dimensional results,We present new results of three-dimensional in vitro cultures of breast cancer cells exhibiting patterns.,683
This patterns. illustrated simulations model result two-dimensional is showing of by the spheroidal,This result is illustrated by two-dimensional simulations of the model showing spheroidal patterns.,684
of in We the in estimation images domain. the problem frequency address motion operating,We address the problem of motion estimation in images operating in the frequency domain.,685
detection is crucial real-life Novelty applications. for,Novelty detection is crucial for real-life applications.,686
"to in While common closed-set is it activity assume recognition i.e. setting, a","While it is common in activity recognition to assume a closed-set setting, i.e.",687
"of samples test is assumption categories, are in real-world scenario. this a training always impractical","test samples are always of training categories, this assumption is impractical in a real-world scenario.",688
be never can training. including seen samples those before during of categories various Test,Test samples can be of various categories including those never seen before during training.,689
"each the portal, tallied For compared and our capabilities. available assessment","For each portal, our assessment compared and tallied the available capabilities.",690
of include strengths its The metabolic and tools. KEGG genomic,The strengths of KEGG include its genomic and metabolic tools.,691
Bacteria Ensembl of number its strengths large genomic of genomes. include and The tools,The strengths of Ensembl Bacteria include its genomic tools and large number of genomes.,692
genomic and its of metabolic tools include strengths KBase The models.,The strengths of KBase include its genomic tools and metabolic models.,693
"existing appealing our RGB-D detection properties. object several to frameworks, has approach Compared","Compared to existing RGB-D object detection frameworks, our approach has several appealing properties.",694
characteristics the the significant also resulting of ED. firing differences find We in,We also find significant differences in the resulting firing characteristics of the ED.,695
('clean' less a If the better is data) K-Means important latter option. is,If the latter is less important ('clean' data) K-Means is a better option.,696
standard a here procedure. suggested become to simple enough The recipe is,The recipe suggested here is simple enough to become a standard procedure.,697
usefulness An evaluation of the Job demonstrates application Corps the the method. to program empirical,An empirical application to the Job Corps program evaluation demonstrates the usefulness of the method.,698
"studies structures. of have relatively to examined network data the FLM with applications few However,","However, relatively few studies have examined the applications of FLM to data with network structures.",699
also utility data employed demonstrate weather the of to A is the real SFLM.,A real weather data is also employed to demonstrate the utility of the SFLM.,700
"handwritten developed task. MOE architecture In character work, alphanumeric prototype for recognition we this a","In this work, we developed a prototype MOE architecture for handwritten alphanumeric character recognition task.",701
digital fault tolerant candidate a is in computing for (SC) circuits. promising computing Stochastic,Stochastic computing (SC) is a promising candidate for fault tolerant computing in digital circuits.,702
tolerance stochastic computing This estimation for utilizing fault algorithms. the of high implementing allows,This allows utilizing the high fault tolerance of stochastic computing for implementing estimation algorithms.,703
presented is The the based proposed recently algorithm. Sparse on Kaczmarz architecture linearized-Bregman-based,The presented architecture is based on the recently proposed linearized-Bregman-based Sparse Kaczmarz algorithm.,704
thermal for The proposed approach is tested images.,The proposed approach is tested for thermal images.,705
"identified local were and Specifically, the Generator cropped lung Lung regions by the (LRG). Region","Specifically, the local lung regions were identified and cropped by the Lung Region Generator (LRG).",706
"were fused for obtained by features module the the fusion Lastly, feature classification. disease","Lastly, the obtained features were fused by the feature fusion module for disease classification.",707
"our method its two and to image applied When strengths. benchmarks, validates usefulness classification","When applied to two image classification benchmarks, our method validates its usefulness and strengths.",708
for digital used Thumbnails images. the widely over as are preview world all a,Thumbnails are widely used all over the world as a preview for digital images.,709
existing indicate performance superior techniques. the experimental of state-of-the-art proposed over model The results the,The experimental results indicate the superior performance of the proposed model over existing state-of-the-art techniques.,710
"Finally, and each at neuronal computed layer are compared. similarities","Finally, neuronal similarities at each layer are computed and compared.",711
improvement significant work. over Results indicate previous,Results indicate significant improvement over previous work.,712
change Such models dynamics neglect how interactions human over of the time. inter-related,Such models neglect the inter-related dynamics of how human interactions change over time.,713
three groups. We evidence estimators can be that categorized into the provide,We provide evidence that the estimators can be categorized into three groups.,714
consistently all first levels. group performs across and DGPs aggregation The well,The first group performs consistently well across all DGPs and aggregation levels.,715
The shows for particular only competitive DGPs. performance group second,The second group shows competitive performance only for particular DGPs.,716
group the is by The estimators. other third clearly outperformed,The third group is clearly outperformed by the other estimators.,717
"high-resolution Regression propose for (FHR) this alignment. a Fractional video-based In paper, we Heatmap face","In this paper, we propose a Fractional Heatmap Regression (FHR) for high-resolution video-based face alignment.",718
distribution accuracy. semantic influence segmentation on a negative Unbalanced have label could,Unbalanced semantic label distribution could have a negative influence on segmentation accuracy.,719
"two contrast reconciles framework Our chromatic opposing and perceptual simultaneous assimilation. phenomena, known as","Our framework reconciles two opposing perceptual phenomena, known as simultaneous contrast and chromatic assimilation.",720
"computing where detection descriptor. the ), differs the precedes phase from and","), where the detection phase precedes and differs from computing the descriptor.",721
diagnostics Next-generation sequencing and bacterial routine technology for pathogens of genetic enables detection clinical research.,Next-generation sequencing technology enables routine detection of bacterial pathogens for clinical diagnostics and genetic research.,722
bacterial analysis of sequencing Whole been has in genome importance epidemiologic of the pathogens.,Whole genome sequencing has been of importance in the epidemiologic analysis of bacterial pathogens.,723
"for are applications. practical genome available sequencing-based few pipelines However, whole genotyping","However, few whole genome sequencing-based genotyping pipelines are available for practical applications.",724
work Papyan et recent by A al.,A recent work by Papyan et al.,725
"SBDL suggested while the patches. algorithm on CSC, operating image the for locally","suggested the SBDL algorithm for the CSC, while operating locally on image patches.",726
"to compared performance the SBDL the on better relying demonstrates ADMM. methods, albeit Fourier-based still","SBDL demonstrates better performance compared to the Fourier-based methods, albeit still relying on the ADMM.",727
have achieved Various been and approaches desirable results. highly proposed,Various approaches have been proposed and achieved highly desirable results.,728
"its paired data Nevertheless, acquire. which abundant to training expensive requires are accomplishment","Nevertheless, its accomplishment requires abundant paired training data which are expensive to acquire.",729
automatically focused generated through data. on Our paired work learning is,Our work is focused on learning through automatically generated paired data.,730
methods network neural have (CNN) for Convolutional image achieved success recently (SR). based super-resolution great,Convolutional neural network (CNN) based methods have recently achieved great success for image super-resolution (SR).,731
"However, attempt models CNN to improve SR most measures deep based (e.g. distortion","However, most deep CNN based SR models attempt to improve distortion measures (e.g.",732
"PSNR, while IFC, quantified (e.g. poor quality VIF) perceptual resulting in SSIM,","PSNR, SSIM, IFC, VIF) while resulting in poor quantified perceptual quality (e.g.",733
"as human such opinion score, measures quality NIQE). no-reference","human opinion score, no-reference quality measures such as NIQE).",734
ineffective partially. either only tackle the methods or The existing are problem,The existing methods either are ineffective or only tackle the problem partially.,735
ii) the Constraints different imposed parametrization on to simultaneously. train be modules can,ii) Constraints can be imposed on the parametrization to train different modules simultaneously.,736
"constraints inference. as joint training the call and ""cooperative enable We these losses"" they","We call these constraints ""cooperative losses"" as they enable the joint training and inference.",737
structure discovery are thematic models used widely for text. Topic in,Topic models are widely used for thematic structure discovery in text.,738
at hand. topic inference procedures often dedicated But traditional tasks models for specific require,But traditional topic models often require dedicated inference procedures for specific tasks at hand.,739
"word-level designed are Also, generate representations. to they semantic not","Also, they are not designed to generate word-level semantic representations.",740
"produce generator Meanwhile, word-level the could also semantic representations.","Meanwhile, the generator could also produce word-level semantic representations.",741
"ATM from articles. news to meaningful able events extract is Moreover,","Moreover, ATM is able to extract meaningful events from news articles.",742
examination conduct empirical of Convolutional seven Network Neural a We (CNN) architectures. thorough,We conduct a thorough empirical examination of seven Convolutional Neural Network (CNN) architectures.,743
Different could under regularization the be terms model. same evaluated,Different regularization terms could be evaluated under the same model.,744
"in degradation the be solved could by out-of-focus speckle setup Moreover, easily. light illumination entailed","Moreover, the degradation entailed by out-of-focus light could be solved in speckle illumination setup easily.",745
the datasets two proposed of effectiveness The demonstrates on the approach. benchmark performance,The performance on two benchmark datasets demonstrates the effectiveness of the proposed approach.,746
"worse, and To color existing matters the real make spaces. overlapped faces are in fake","To make matters worse, the real and fake faces are overlapped in existing color spaces.",747
videos. dependencies and is capturing designed in The spatio-temporal long-range images non-local for module,The non-local module is designed for capturing long-range spatio-temporal dependencies in images and videos.,748
"within optimization. to generalized we method ease the implement channel our groups non-local Moreover,","Moreover, we implement our generalized non-local method within channel groups to ease the optimization.",749
Audio clips. labels descriptive from infer aims tagging audio to,Audio tagging aims to infer descriptive labels from audio clips.,750
Audio of limited is size tagging challenging data due and labels. the to noisy,Audio tagging is challenging due to the limited size of data and noisy labels.,751
the to training strategy employed ensemble address for label sample problem. re-weight is a noisy,a sample re-weight strategy is employed for ensemble training to address the noisy label problem.,752
Science. inference is Network problem a prominent Network topology in,Network topology inference is a prominent problem in Network Science.,753
"networks, cost. pruning to trained redundant unstructured weights can storage deep lower reduce In neural","In trained deep neural networks, unstructured pruning can reduce redundant weights to lower storage cost.",754
"requires hardwares the it speed However, practical up inference. to of customization","However, it requires the customization of hardwares to speed up practical inference.",755
sacrifices this But method often model accuracy.,But this method often sacrifices model accuracy.,756
generation challenging yet Caricature interesting an is task.,Caricature generation is an interesting yet challenging task.,757
is plausible reasonable exaggerations The face to with goal primary generate given caricatures images.,The primary goal is to generate plausible caricatures with reasonable exaggerations given face images.,758
"data usage Moreover, models which limits largely paired some scenarios. require training their strictly","Moreover, some models require strictly paired training data which largely limits their usage scenarios.",759
"paper, this these propose problems. In CariGAN overcome we","In this paper, we propose CariGAN overcome these problems.",760
"data, learns CariGAN only transformations images. on of from paired paired weakly training Instead","Instead of training on paired data, CariGAN learns transformations only from weakly paired images.",761
chirp al. presented et The algorithm Howard originally by detection,The chirp detection algorithm originally presented by Howard et al.,762
into and integrated slotting peeling a framework. is for decoder patching designed and enhanced a,is enhanced and integrated into a peeling decoder designed for a patching and slotting framework.,763
of is A advantage efficiency. computational its significant algorithm our,A significant advantage of our algorithm is its computational efficiency.,764
"only dataset. the from After we KITTI sub-sequences evaluation, our kept","After our evaluation, we kept only sub-sequences from the KITTI dataset.",765
vehicle. data collection sequences will In future of our from increase the with the we,In the future we will increase the collection of sequences with data from our vehicle.,766
other outperforms that DUNet indicates Results state-of-the-arts.,Results indicates that DUNet outperforms other state-of-the-arts.,767
tone. transmission light Each multiplexed unit with unique is a,Each light unit transmission is multiplexed with a unique tone.,768
consider virus capsid the by molecular We methods of of into dynamics. proteins self-assembly a,We consider self-assembly of proteins into a virus capsid by the methods of molecular dynamics.,769
augmented proteins are and the by coarse-grained structure-based The described interactions. by electrostatic flexible model,The proteins are flexible and described by the structure-based coarse-grained model augmented by electrostatic interactions.,770
"supramolecular involved Previous solid e.g. the of self-assembly a studies capsid objects scale, of","Previous studies of the capsid self-assembly involved solid objects of a supramolecular scale, e.g.",771
"and couplings engineered to with corresponding stochastic capsomeres, movements.","corresponding to capsomeres, with engineered couplings and stochastic movements.",772
results show architecture our The that methods. other state-of-the-art outperforms,The results show that our architecture outperforms other state-of-the-art methods.,773
when handcrafted has Learning to detection detection. feature unexplored an largely feature area been compared,Learning feature detection has been largely an unexplored area when compared to handcrafted feature detection.,774
covariant function covariant loss use constraint their the Recent learning detectors. learn in formulations to,Recent learning formulations use the covariant constraint in their loss function to learn covariant detectors.,775
"just from to learning detection unstable lead features. constraint covariant However, of can","However, just learning from covariant constraint can lead to detection of unstable features.",776
"further, trained hand-crafted To to impart stability by detectors are detectors. pre-determined obtained features extract","To impart further, stability detectors are trained to extract pre-determined features obtained by hand-crafted detectors.",777
"in However, to they the process the lose features. detect novel ability","However, in the process they lose the ability to detect novel features.",778
image problem critical (SR) Image quality is algorithms. super-resolution a for measurement,Image quality measurement is a critical problem for image super-resolution (SR) algorithms.,779
ophthalmologists make to Fundoscopic often investigated by abnormal lesions are spot images to diagnoses.,Fundoscopic images are often investigated by ophthalmologists to spot abnormal lesions to make diagnoses.,780
Generative two Adversarial We techniques new (GANs). propose training Networks for,We propose two new techniques for training Generative Adversarial Networks (GANs).,781
inverse t-SNE to We propose this. regularizer achieve an,We propose an inverse t-SNE regularizer to achieve this.,782
between difference the discriminator of We scores constrain generated the the samples and real ones.,We constrain the difference between the discriminator scores of the real samples and generated ones.,783
gradients the further scores. We of these difference the between constrain discriminator,We further constrain the difference between the gradients of these discriminator scores.,784
derive of Taylor We discriminator approximations the these function. from constraints,We derive these constraints from Taylor approximations of the discriminator function.,785
under impoverished synchronization networks harsh conditions. concerns the paper sensor infrastructure of This with,This paper concerns with the synchronization of infrastructure impoverished sensor networks under harsh conditions.,786
"three asynchronous, suggests time synchronization It energyefficient protocols. decentralized, novel","It suggests three novel asynchronous, decentralized, energyefficient time synchronization protocols.",787
the protocols. analysis well these protocols known outperform that the All show,All the analysis show that these protocols outperform the well known protocols.,788
"realistic are the ones. they being more asynchronous, to synchronous relative Also,","Also, being asynchronous, they are more realistic relative to the synchronous ones.",789
neural organization. how code raises of processes of This issue the these further the supports,This raises the further issue of how the neural code supports these processes of organization.,790
"the However, lack an task. robots robotic in of makes exceptionally challenging intuition grasping such","However, the lack of such intuition in robots makes robotic grasping an exceptionally challenging task.",791
grasping equally of object. often an There options several viable are,There are often several equally viable options of grasping an object.,792
"second According official to of as statistics, cancer human is leading cause the considered fatalities.","According to official statistics, cancer is considered as the second leading cause of human fatalities.",793
"recent our work, developed overcome shortcoming. to In CapsNets newly have incorporated this we","In our recent work, we have incorporated newly developed CapsNets to overcome this shortcoming.",794
"to CapsNets image background. sensitive highly are, miscellaneous the however,","CapsNets are, however, highly sensitive to the miscellaneous image background.",795
approach its outperforms noticeably proposed counterparts. The,The proposed approach noticeably outperforms its counterparts.,796
with based Resampling is it because flow disocclusions. on deal cannot insufficient,Resampling based on flow is insufficient because it cannot deal with disocclusions.,797
lead results. models blurry Generative currently to,Generative models currently lead to blurry results.,798
Recent a by patches predicted input convolving with pixel synthesis a approaches kernel.,Recent approaches synthesis a pixel by convolving input patches with a predicted kernel.,799
"However, increases requirement their size. memory with kernel","However, their memory requirement increases with kernel size.",800
"spatially-displaced we (SDC) module Here, video frame prediction. convolution for","Here, we spatially-displaced convolution (SDC) module for video frame prediction.",801
motion Our frames effectively consistent crisp and motion. handles model large synthesizes with,Our model handles large motion effectively and synthesizes crisp frames with consistent motion.,802
requires Accurately with knowledge. a combining about answering question a general image given observations,Accurately answering a question about a given image requires combining observations with general knowledge.,803
"with general humans, an this remains reasoning algorithmic is knowledge challenge. While for effortless","While this is effortless for humans, reasoning with general knowledge remains an algorithmic challenge.",804
a to Bi-Real two training net We challenges. novel with propose these algorithm tackle,We propose Bi-Real net with a novel training algorithm to tackle these two challenges.,805
if results treated early. loss Diabetic retinopathy in vision (DR) not,Diabetic retinopathy (DR) results in vision loss if not treated early.,806
been on techniques have traditional based machine-learning (ML) Many hand-engineered introduced. features,Many traditional machine-learning (ML) techniques based on hand-engineered features have been introduced.,807
"highlighting review this In their pros and these cons. paper, methods, we","In this paper, we review these methods, highlighting their pros and cons.",808
validated Short-term against case model counts. projections actual were,Short-term model projections were validated against actual case counts.,809
"forecasts. projections, consistently higher on models short-term of validation shorter-term During scored","During validation of short-term projections, models consistently scored higher on shorter-term forecasts.",810
identification partial models nonseparable develop approach in endogeneity. for effects I with new a,I develop a new identification approach for partial effects in nonseparable models with endogeneity.,811
authentication the provide Mobile an approaches convenience biometric with of technology. omnipresent secure,Mobile biometric approaches provide the convenience of secure authentication with an omnipresent technology.,812
"this tuning focuses Particularly, matrix of adaptive the study fly. the weighting on a on","Particularly, this study focuses on the adaptive tuning of a weighting matrix on the fly.",813
CNN with Classification as pre-trained features results features. were as AE discriminative,Classification results with AE features were as discriminative as pre-trained CNN features.,814
SEQR SEQuence We Retrieval. for this method call,We call this method SEQR for SEQuence Retrieval.,815
"transmission approaches energy. edge low and computing consume data reduce Alternatively,","Alternatively, edge computing approaches reduce data transmission and consume low energy.",816
for WDM user. high a utilized data rate to is each attain,WDM is utilized to attain a high data rate for each user.,817
intensity Conventional cameras information. reference low-frequency capture,Conventional cameras capture low-frequency reference intensity information.,818
two sensor These information. complementary modalities provide,These two sensor modalities provide complementary information.,819
"only. be the can events of conventional absence image run frames, on filter In","In absence of conventional image frames, the filter can be run on events only.",820
"arduous placed on data An is, and hence, design burden the acquisition. framework learning","An arduous burden is, hence, placed on the learning framework design and data acquisition.",821
called characterizing system approach mapping is identification. the A stimulus-response general function for,A general approach for characterizing the stimulus-response mapping function is called system identification.,822
data different the the how makes about algorithm neuron Each and generated. assumptions is,Each algorithm makes different assumptions about the neuron and how the data is generated.,823
has C++ Calling code its perils. from multi-threaded R,Calling multi-threaded C++ code from R has its perils.,824
"R synchronize thread. with main from One can, the however,","One can, however, synchronize with R from the main thread.",825
apply mechanism to also synchronization other frameworks. threading may The,The synchronization mechanism may also apply to other threading frameworks.,826
"models. no type been package can fit this Until published, now, which of R has","Until now, no R package has been published, which can fit this type of models.",827
"Meanwhile, eLIAN the of nearly the LIAN. the as quality of is same solution one","Meanwhile, the solution quality of eLIAN is nearly the same as the one of LIAN.",828
with in subregions elevation ten were ranges. comparable replicated Analyses,Analyses were replicated in ten subregions with comparable elevation ranges.,829
infrastructureless for is Cooperative a wireless using vehicles (UAVs) technology unmanned communication promising networks. aerial,Cooperative communication using unmanned aerial vehicles (UAVs) is a promising technology for infrastructureless wireless networks.,830
of in throughput. backhaul communications the based challenges UAV One key the is,One of the key challenges in UAV based communications is the backhaul throughput.,831
"investigate properties e.g., estimators, important these and We efficiency. of bias","We investigate important properties of these estimators, e.g., bias and efficiency.",832
through also We the explore performance estimators numerical these examples. of,We also explore the performance of these estimators through numerical examples.,833
light low noise short-exposure conditions. frames image reduced with in an provide can Merging,Merging short-exposure frames can provide an image with reduced noise in low light conditions.,834
"to is problem. open images However, best align how before an merging","However, how best to align images before merging is an open problem.",835
to initial form matrix translations guess are rotation homography. The and of combined the,The rotation matrix and translations are combined to form the initial guess of homography.,836
is An more filter homography utilized unscented to a accurate estimation. Kalman provide,An unscented Kalman filter is utilized to provide a more accurate homography estimation.,837
We the proposed with results. compare and benchmark methods to single-image multi-image favorable denoising method,We compare the proposed method to benchmark single-image and multi-image denoising methods with favorable results.,838
approach of integrate-and-fire. means This is leaky by exemplified,This approach is exemplified by means of leaky integrate-and-fire.,839
"present a approach we this the In stereo to photometric multi-view volumetric work, problem.","In this work, we present a volumetric approach to the multi-view photometric stereo problem.",840
real and evaluated and is achieves state-of-the-art results. on Our synthetic data-sets approach,Our approach is evaluated on synthetic and real data-sets and achieves state-of-the-art results.,841
kernel Gaussian approximating Fourier of truncated (trigonometric) using a problem consider the We functions.,We consider the problem of approximating a truncated Gaussian kernel using Fourier (trigonometric) functions.,842
This fields. many processing method applicable image decomposition to is,This decomposition method is applicable to many image processing fields.,843
"image In the paper, fusion we focus this on task.","In this paper, we focus on the image fusion task.",844
real-world show opening We lock. a an example,We show an example opening a real-world lock.,845
"area Noise, (like etc.) Electrical/Device Threshold, properties","Electrical/Device properties (like Noise, Threshold, area etc.)",846
circuit in also the been discussed of this proposed work. have,of the proposed circuit have also been discussed in this work.,847
"Generally, convolutional data a on regular networks process grid, (CNNs) e.g. neural","Generally, convolutional neural networks (CNNs) process data on a regular grid, e.g.",848
is transportation key Driving understanding intelligent for Scene a ingredient systems.,Driving Scene understanding is a key ingredient for intelligent transportation systems.,849
with We driving datasets. other to a of comparison detailed provide analysis a HDD,We provide a detailed analysis of HDD with a comparison to other driving datasets.,850
"the proposed of the effectiveness we paper, using In a measure retrieval this experiment. evaluate","In this paper, we evaluate the effectiveness of the proposed measure using a retrieval experiment.",851
has proved NP-complete Finding optimal a been order the problem. variable,Finding the optimal variable order has been proved a NP-complete problem.,852
near-optimal this Many proposed solution have a find been heuristics problem. to of,Many heuristics have been proposed to find a near-optimal solution of this problem.,853
extension performance applications. This beamforming enables distributed range for,This performance enables distributed beamforming for range extension applications.,854
"Deep RNN, classification, LSTM, QRS Keywords: detection. learning, ECG Bi-directional","Keywords: ECG classification, Deep learning, RNN, Bi-directional LSTM, QRS detection.",855
"over-burden the potential explosion data and storage likely communication to is systems. However,","However, the potential data explosion is likely to over-burden storage and communication systems.",856
case evaluated our We using studies. approach have two,We have evaluated our approach using two case studies.,857
are explored regarding our variables in performances several The detail. approach BSM recovery of,The recovery performances of our approach regarding several BSM variables are explored in detail.,858
of the hardware OBU cost. show amounts potential results of The large saving,The results show the potential of saving large amounts of OBU hardware cost.,859
Person challenging are available (re-ID) problem no re-identification for labels when is a especially training.,Person re-identification (re-ID) is a challenging problem especially when no labels are available for training.,860
includes model. generation components: Our of virtual approach and training deep person two re-ID,Our approach includes two components: virtual person generation and training of deep re-ID model.,861
images. bilateral with nonlocal grayscale means work Existing fast filtering for and mostly algorithms,Existing fast algorithms for bilateral and nonlocal means filtering mostly work with grayscale images.,862
"fast bilateral In propose filtering. and for nonlocal a high-dimensional means paper, algorithm we this","In this paper, we propose a fast algorithm for high-dimensional bilateral and nonlocal means filtering.",863
"its intense bilateral the Similar brute-force the to of requires implementation counterpart filter, computations. adaptive","Similar to the bilateral filter, the brute-force implementation of its adaptive counterpart requires intense computations.",864
"the texture deblocking, demonstrate our and JPEG for algorithm We of effectiveness sharpening, filtering.","We demonstrate the effectiveness of our algorithm for sharpening, JPEG deblocking, and texture filtering.",865
observable partially systems. for fully and present We formulations observable,We present formulations for fully observable and partially observable systems.,866
"previous hallucination recovery. identity facial largely approaches However, face ignore","However, previous face hallucination approaches largely ignore facial identity recovery.",867
"faces. of significantly improves SICNN addition, the ultra-low-resolution recognizability In","In addition, SICNN significantly improves the recognizability of ultra-low-resolution faces.",868
object many Video advanced fundamental segmentation vision in applications. step is a,Video object segmentation is a fundamental step in many advanced vision applications.,869
"in the similar generates test sample. backgrounds to the testing, GAN trained During that","During testing, the trained GAN generates backgrounds similar to that in the test sample.",870
moving compared with The generated are background samples sample then objects. the to test segment,The generated background samples are then compared with the test sample to segment moving objects.,871
"and is the layer evaluated Similarly, ComposeCaps baseline. improvement the over demonstrates an","Similarly, the ComposeCaps layer is evaluated and demonstrates an improvement over the baseline.",872
"to compresses convolutional multi-level Specifically, proposed features. spatial DSNet uniform features the resolution","Specifically, the proposed DSNet compresses multi-level convolutional features to uniform spatial resolution features.",873
sociology. philosophy and the models explanations between distinction these in We and on focus,We focus on the distinction between these models and explanations in philosophy and sociology.,874
"the started lung CT be screening world. to has Therefore, used widely across all","Therefore, CT lung screening has started to be used widely all across the world.",875
"serious However, images radiologists. a burden for analyzing these is","However, analyzing these images is a serious burden for radiologists.",876
used class-imbalance. to Mini-batch overcome is in order,Mini-batch is used in order to overcome class-imbalance.,877
WaveFlow. can available and algorithms data using in real-time in both Raw be processed TensorFlow,Raw data can be processed in real-time using algorithms available both in TensorFlow and WaveFlow.,878
"B-mode processing and provides for reconstruction ops image ultrasound. (beamforming), signal WaveFlow quantitative Currently,","Currently, WaveFlow provides ops for B-mode image reconstruction (beamforming), signal processing and quantitative ultrasound.",879
WaveFlow source be github.com/waveflow-team/waveflow at found can . code,WaveFlow source code can be found at github.com/waveflow-team/waveflow .,880
metric this canonical characterizes for sampling. Hermann measure discrepancy threshold-based This paper making structure Weyl's,This paper characterizes this metric structure making Hermann Weyl's discrepancy measure canonical for threshold-based sampling.,881
"spatial-temporal on maps. we global convolution apply temporal To model local relationship, feature spatial-temporal the","To model global spatial-temporal relationship, we apply temporal convolution on the local spatial-temporal feature maps.",882
"temporal novel Xception StNet. in Specifically, is block a proposed","Specifically, a novel temporal Xception block is proposed in StNet.",883
convolution It sequence separate video. and channel-wise temporal-wise feature the over employs of a,It employs a separate channel-wise and temporal-wise convolution over the feature sequence of video.,884
realistic convergence of power for challenging. scalable be simulation Achieving cases can robust and flow,Achieving robust and scalable convergence for simulation of realistic power flow cases can be challenging.,885
Table The to the reader I refer summary can for formulas. of these,The reader can refer to Table I for the summary of these formulas.,886
"successfully recognition the learning micro-expression deep have area. Recently, methods introduced been into","Recently, deep learning methods have been successfully introduced into the micro-expression recognition area.",887
"substantial challenges accuracy recognition in the micro-expression remain. achieved, recognition Whilst higher","Whilst the higher recognition accuracy achieved, substantial challenges in micro-expression recognition remain.",888
"Recently, estimation antenna considered reduced with arrays hardware for have DoA been compressive complexity.","Recently, compressive antenna arrays have been considered for DoA estimation with reduced hardware complexity.",889
on descent design gradient (SGD) stochastic is proposed grid-free The optimization. based,The proposed design is based on grid-free stochastic gradient descent (SGD) optimization.,890
"recognition is a research research area, In problem. hot expression facial computer the","In the computer research area, facial expression recognition is a hot research problem.",891
"circumstances. research has lab the years, from the to environment moved in-the-wild Recent","Recent years, the research has moved from the lab environment to in-the-wild circumstances.",892
"It poses. especially under challenging, is extreme","It is challenging, especially under extreme poses.",893
"In we the solve problem work, approach. this in opposite the","In this work, we solve the problem in the opposite approach.",894
poses special head detect expressions consider within head We the poses. and the,We consider the head poses and detect the expressions within special head poses.,895
"implement comparison, the forest SVM models. a prediction we as and to As random","As a comparison, we implement SVM and random forest to as the prediction models.",896
networks found of dispersal to structures networks parasite and identical. seed host The are be,The structures of host parasite networks and seed dispersal networks are found to be identical.,897
"To this this tackle issue, Learning Metric Point-based Distance paper. in (PDML) a we propose","To tackle this issue, we propose a Point-based Distance Metric Learning (PDML) in this paper.",898
data methods large detection training with every logo annotations class. general-purpose for require object Existing,Existing general-purpose object detection methods require large training data with annotations for every logo class.,899
The of branch. conditional a consists proposed segmentation branch and model a,The proposed model consists of a conditional branch and a segmentation branch.,900
"protocol i.e. the MIMO optimization radar, analyse We that RL-based allows an","We analyse an RL-based optimization protocol that allows the MIMO radar, i.e.",901
"sense \textit{agent}, the to iteratively i.e. environment, unknown the","the \textit{agent}, to iteratively sense the unknown environment, i.e.",902
FC. harvesting energy of communication storing capable and for Each sensor is the with,Each sensor is capable of harvesting and storing energy for communication with the FC.,903
that indicate are from results maximizing {\theta}k's obtained distance Our near-optimal. numerical the KL,Our numerical results indicate that {\theta}k's obtained from maximizing the KL distance are near-optimal.,904
stereo Photometric a assumes illumination illuminated only source. that scene directly is the by,Photometric stereo assumes that a scene is illuminated only directly by the illumination source.,905
"in effects biases shape. the recovered illumination result, indirect introduce strong inter-reflections As due to","As result, indirect illumination effects due to inter-reflections introduce strong biases in the recovered shape.",906
approach recover scene to is presence indirect illumination. properties suggested of the in Our,Our suggested approach is to recover scene properties in the presence of indirect illumination.,907
"of They however, points snapshot distances. from serve localize to static objects a which are,","They are, however, static objects which serve to localize points from a snapshot of distances.",908
"If objects do motion. the expects modeling one to by the better move,","If the objects move, one expects to do better by modeling the motion.",909
"time, squared of become time-varying functions the entries of The KEDMs distances.","The entries of KEDMs become functions of time, the squared time-varying distances.",910
"inspired a contribution SDRs EDMs. (SDR), relaxation is by main Our for static semidefinite","Our main contribution is a semidefinite relaxation (SDR), inspired by SDRs for static EDMs.",911
problems geometry This makes in for KEDMs dynamic of points a sets. promising tool,This makes KEDMs a promising tool for problems in geometry of dynamic points sets.,912
density information. fully single designed utilize not Previous or for pan-density methods do are scenes,Previous methods are designed for single density scenes or do not fully utilize pan-density information.,913
"counting. a pan-density Pan-Density (PaDNet), framework, Network propose for crowd novel the We","We propose a novel framework, the Pan-Density Network (PaDNet), for pan-density crowd counting.",914
(ICA) Coronary the (CAD) gold in Invasive Artery imaging. angiography standard Disease is coronary,Invasive coronary angiography (ICA) is the gold standard in Coronary Artery Disease (CAD) imaging.,915
for labels each cardiac DNN second frame. predicts phase A,A second DNN predicts cardiac phase labels for each frame.,916
"state, or were interventions, related exclusion No formulated. patient criteria previous pathology to","No exclusion criteria related to patient state, previous interventions, or pathology were formulated.",917
to have. imaging range is sensors a that This due dynamic limited,This is due to a limited dynamic range that imaging sensors have.,918
"contrast various been this proposed. enhancement methods For reason, have","For this reason, various contrast enhancement methods have been proposed.",919
function is over-enhancement. preventing used shadow-up for A,A shadow-up function is used for preventing over-enhancement.,920
Image fundamental many graphics and disparate computer smoothing applications. vision of component a represents,Image smoothing represents a fundamental component of many disparate computer vision and graphics applications.,921
"with are GFDM filter, prototype a single on e.g. based Some","Some are based on GFDM with a single prototype filter, e.g.",922
"Accordingly, current consume the resources. high state art of implementations the","Accordingly, the current state of the art implementations consume high resources.",923
by general developing gap address model a this We intelligence. for,We address this gap by developing a model for general intelligence.,924
"accomplish focus aspects three basic of on intelligence. we this, To","To accomplish this, we focus on three basic aspects of intelligence.",925
"we must of level. intelligence high nature at general a and realize the First, order","First, we must realize the general order and nature of intelligence at a high level.",926
"as we realizations as Third, must possible. these describe clearly","Third, we must describe these realizations as clearly as possible.",927
the and a order We capture to propose hierarchical exploit help within model intelligence.,We propose a hierarchical model to help capture and exploit the order within intelligence.,928
useful econometrics. in Finite are mixture applied models,Finite mixture models are useful in applied econometrics.,929
contaminated multiple models in also dealing with equilibria. with are models sampling convenient Mixtures and,Mixtures are also convenient in dealing with contaminated sampling models and models with multiple equilibria.,930
power utilize to covariates implied variation. information The identification in key is the by,The key is to utilize the identification power implied by information in covariates variation.,931
"under of conditions. presented, identification and sufficient approaches three are First, sets non-nested distinct","First, three identification approaches are presented, under distinct and non-nested sets of sufficient conditions.",932
Observable the features is valid. data us three approaches of inform of which,Observable features of data inform us which of the three approaches is valid.,933
"Second, results the extensions identification developed. of some are","Second, some extensions of the identification results are developed.",934
aims instances into classify ordinal regression categories. Ordinal to,Ordinal regression aims to classify instances into ordinal categories.,935
show BMI outperforms results that classification regression-based Experimental and categories. our at methods approach estimating,Experimental results show that our approach outperforms classification and regression-based methods at estimating BMI categories.,936
of illustrated architecture The performance been results. simulation has this also by new,The performance of this new architecture has also been illustrated by simulation results.,937
human of life central the and to longevity population are Estimating dynamics. expectancy the computing,Estimating the human longevity and computing of life expectancy are central to the population dynamics.,938
"including astronomer Edmund Halley. since studied by scientists renowned aspects were seriously These fifteenth century,","These aspects were studied seriously by scientists since fifteenth century, including renowned astronomer Edmund Halley.",939
smoothness existing of preserve use optical to the synthesized video. flow Many temporal methods the,Many existing methods use optical flow to preserve the temporal smoothness of the synthesized video.,940
"of optical rapid flow sensitive and estimation is motions. the occlusions However, to","However, the estimation of optical flow is sensitive to occlusions and rapid motions.",941
sensing computational Compressive for systems. imaging modern technology is a new,Compressive sensing is a new technology for modern computational imaging systems.,942
compressive system linear a with of sensing nonlinear estimate We the model.,We estimate the nonlinear system of compressive sensing with a linear model.,943
research Simulation available development codes are and for purposes.,Simulation codes are available for research and development purposes.,944
"methods a rarely analyzed are Although in filter functionality. these of effective, perspective","Although effective, these methods are rarely analyzed in a perspective of filter functionality.",945
"camera advantages In paper, we such of investigate this the designs.","In this paper, we investigate the advantages of such camera designs.",946
a components the single to separate using specular allows and us diffuse image. essentially This,This essentially allows us to separate the diffuse and specular components using a single image.,947
separation an to by vector. photograph a encoder object performs style mapping latent network The,The encoder network performs style separation by mapping an object photograph to a latent vector.,948
"vector halves latent represent respectively. and content style, the Two the of","Two halves of the latent vector represent the content and style, respectively.",949
performs The taking as generator concatenated vector input. a synthesis style by network,The generator network performs style synthesis by taking a concatenated vector as input.,950
derived finally GBI resulting the is measured saliency integrating The of by geometric buildings.,The resulting GBI is finally measured by integrating the derived geometric saliency of buildings.,951
selection automatic sorting. online channels meaningful of offline spike during and,automatic selection of meaningful channels during offline and online spike sorting.,952
address to enables issues. The USC us both above-raised,The USC enables us to address both above-raised issues.,953
(Conv convolutional net). the standard USC of The neural uses networks architecture,The USC uses the standard architecture of convolutional neural networks (Conv net).,954
the from Hence detailed lack geometry. a suffers of method,Hence the method suffers from a lack of detailed geometry.,955
laptop. few on a minutes take standard The calculations a,The calculations take a few minutes on a standard laptop.,956
popular neural one (CoNNs) most belong of DNN of the networks Convolutional to types architectures.,Convolutional neural networks (CoNNs) belong to one of the most popular types of DNN architectures.,957
for paper and presents evaluation design This an of the CoNNs. accelerator,This paper presents the design and evaluation of an accelerator for CoNNs.,958
"is cellular The based (CeNNs). mixed-signal, neural system-level architecture networks on","The system-level architecture is based on mixed-signal, cellular neural networks (CeNNs).",959
"of and outperforms Moreover, approaches terms in it the robustness. efficiency, accuracy, state-of-the-art","Moreover, it outperforms the state-of-the-art approaches in terms of accuracy, efficiency, and robustness.",960
(KAF) is detection. filtering direct optical nonlinearity-tolerant Kernel for proposed adaptive,Kernel adaptive filtering (KAF) is proposed for nonlinearity-tolerant optical direct detection.,961
its in goals. to actions order agent achieve An performs intelligent,An intelligent agent performs actions in order to achieve its goals.,962
different values: computation of two We present dynamic. and static definitions,We present two different definitions of computation values: static and dynamic.,963
"temporal that consistency asymptotic desirable convergence. and show as values properties, such have We these","We show that these values have desirable properties, such as temporal consistency and asymptotic convergence.",964
"methods static computing the and propose dynamic approximating we Furthermore, and computation efficiently values. for","Furthermore, we propose methods for efficiently computing and approximating the static and dynamic computation values.",965
element an of surgical is important Surgical education. increasingly simulation,Surgical simulation is an increasingly important element of surgical education.,966
"adequately natural representation multidimensinal the Moreover, existing approaches to explore of HIs. the fail","Moreover, the existing approaches fail to adequately explore the natural multidimensinal representation of HIs.",967
"solution Unfortunately, model aggregates. is help structurally to no there existing sound","Unfortunately, there is no existing solution to help model structurally sound aggregates.",968
of external This attachment done specifying set a is forces points. by and,This is done by specifying a set of external forces and attachment points.,969
The specified a that scenario. automatically elements the user-provided then layout of algorithm optimizes answers,The algorithm then automatically optimizes a layout of user-provided elements that answers the specified scenario.,970
"elongated, be deform. shapes: and The to allowed convex, concave, can arbitrary can have elements","The elements can have arbitrary shapes: convex, concave, elongated, and can be allowed to deform.",971
descent gradient through scheme. a maximized Global dedicated is rigidity,Global rigidity is maximized through a dedicated gradient descent scheme.,972
(GSP). in Graph graph processing with signal a problem sampling fundamental noise is,Graph sampling with noise is a fundamental problem in graph signal processing (GSP).,973
"minimize condition matrix. }, the of number the coefficient","}, minimize the condition number of the coefficient matrix.",974
$T$ sample $K$. We perform to maximize binary given then a search budget,We then perform binary search to maximize $T$ given a sample budget $K$.,975
predictions our shows well. perform networks that neural are able experiment to The trained,The experiment shows that our trained neural networks are able to perform predictions well.,976
method magnitude a determine uses The unique methodology threshold. to bin a contrario,The method uses a contrario methodology to determine a unique bin magnitude threshold.,977
"failed. of Contrast invariance, special subspace a invariance, also case","Contrast invariance, a special case of subspace invariance, also failed.",978
"contrast, some with while increased. neurons of The responses increasing others decreased","The responses of some neurons decreased with increasing contrast, while others increased.",979
"population generating great coarse fine-scaled population of data is significance. from Therefore, distributions","Therefore, generating fine-scaled population distributions from coarse population data is of great significance.",980
a from on real-life collected We dataset experiments extensive Shanghai. perform mobile,We perform extensive experiments on a real-life mobile dataset collected from Shanghai.,981
auxiliary model. after of our The runtime the without training discarded affecting networks efficiency are,The auxiliary networks are discarded after training without affecting the runtime efficiency of our model.,982
Berkeley (BDD) shown are results Encouraging Drive Deep on also dataset.,Encouraging results are also shown on Berkeley Deep Drive (BDD) dataset.,983
concerned frequencies is multi-dimensional paper from estimating unknown with measurements. linear compressive This,This paper is concerned with estimating unknown multi-dimensional frequencies from linear compressive measurements.,984
theory steps. developed three in is The,The theory is developed in three steps.,985
"theory these the them. paper semantics, The fixpoint results existence for of investigates establishing","The paper investigates the fixpoint theory of these semantics, establishing existence results for them.",986
"in compared with Finally, theory the approaches. detail is existing","Finally, the theory is compared in detail with existing approaches.",987
"are wrong updates tracking susceptible correlation models to inaccurate stemming results. based filter However, from","However, correlation filter based models are susceptible to wrong updates stemming from inaccurate tracking results.",988
"update date, has correlation effort towards little To the handling been devoted problem. filter","To date, little effort has been devoted towards handling the correlation filter update problem.",989
"we propose approach a novel In filter address update problem. this paper, the correlation to","In this paper, we propose a novel approach to address the correlation filter update problem.",990
tackles of grayscale colorizing the challenge images. paper This,This paper tackles the challenge of colorizing grayscale images.,991
"show colorization. task extending of We our to application model, video the an","We show an application of our model, extending the task to video colorization.",992
and of experiments Calibration effectiveness real-world to Wizard. performed demonstrate are Synthetic the,Synthetic and real-world experiments are performed to demonstrate the effectiveness of Calibration Wizard.,993
"does During to patches. it hide not any need testing,","During testing, it does not need to hide any patches.",994
"scheme this segmentation. propose new we a pre-training image for based paper, U-net In","In this paper, we propose a new pre-training scheme for U-net based image segmentation.",995
increasingly being diverse sensors many Photon are counting used in applications.,Photon counting sensors are being increasingly used in many diverse applications.,996
information spatial effectively present measurements to in combines reduce Our temporal ConvNet-based scheme and noise.,Our ConvNet-based scheme effectively combines spatial and temporal information present in measurements to reduce noise.,997
"we Based flow information classes. paths different on findings, the reveal for of critical such","Based on such findings, we reveal the critical paths of information flow for different classes.",998
detect image apical slices. to then are These the representations basal used and missing,These image representations are then used to detect the missing basal and apical slices.,999
"for However, labels are learning human inappropriate models. annotated sometimes","However, human annotated labels are sometimes inappropriate for learning models.",1000
observed modeling and process models series Markov time powerful (POMP) Partially are for tools analysis.,Partially observed Markov process (POMP) models are powerful tools for time series modeling and analysis.,1001
"on paper, toy two we based show utility problems. of In this methodologies the these","In this paper, we show the utility of these methodologies based on two toy problems.",1002
"under background the of Meanwhile, algorithm the the strengthens of region. robustness it condition complex","Meanwhile, it strengthens the robustness of algorithm under the condition of the complex background region.",1003
This presents of models. structural casual using paper contrastive a model explanation,This paper presents a model of contrastive explanation using structural casual models.,1004
model on causal is models. One based general structural,One general model is based on structural causal models.,1005
important local information the and for in understanding region shape is contextual analysis. Exploring,Exploring contextual information in the local region is important for shape understanding and analysis.,1006
regions. ways of employ contextual or Existing local hand-crafted often studies information explicit to encode,Existing studies often employ hand-crafted or explicit ways to encode contextual information of local regions.,1007
task identification effects pollutant of The important environmental health. an is in,The identification of pollutant effects is an important task in environmental health.,1008
effects formulas. show new the are Experiments done to of,Experiments are done to show the effects of new formulas.,1009
fusion of the information). low-level vs between high-level,between the fusion of low-level vs high-level information).,1010
"established such, equitably As circuit an design important element. uniqueness their conveyors as current have","As such, current conveyors have equitably established their uniqueness as an important circuit design element.",1011
outlines low of and design. the CCCII on works power low voltage,CCCII works on the outlines of low power and low voltage design.,1012
Tunable well signal in as voltage in amplifiers as analog processing applications. find use digital,Tunable voltage amplifiers find use in analog as well as in digital signal processing applications.,1013
dictionary of is learned various basis a image The composed natural from database.,The dictionary is composed of various basis learned from a natural image database.,1014
"reacting because are to highly and military enemy's keep demands time-varying stochastic actions. Moreover, commanders","Moreover, military demands are time-varying and highly stochastic because commanders keep reacting to enemy's actions.",1015
has community. the recently the programming complex interest attracted of Metamodeling mathematical numerical systems of,Metamodeling of complex numerical systems has recently attracted the interest of the mathematical programming community.,1016
simulators time Characterizing is stochastic the more consuming. even,Characterizing the stochastic simulators is even more time consuming.,1017
for cloud We point algorithm a segmentation. novel present,We present a novel algorithm for point cloud segmentation.,1018
"to This parameters, without capacity improves robust more increasing results. leading segmentation expressive the","This improves the expressive capacity without increasing parameters, leading to more robust segmentation results.",1019
method. both on demonstrate real our synthetic superiority experiments Comprehensive the of and datasets,Comprehensive experiments on both synthetic and real datasets demonstrate the superiority of our method.,1020
among a are leading Hip adults. older of fractures disability and cause death,Hip fractures are a leading cause of death and disability among older adults.,1021
fractures diagnosis pelvic on commonly missed the Hip are radiographs. also most,Hip fractures are also the most commonly missed diagnosis on pelvic radiographs.,1022
and anatomical is visualizing brain. the quantifying Brain of for used segmentation structures image,Brain image segmentation is used for visualizing and quantifying anatomical structures of the brain.,1023
that proposed shows Experimental the outperforms traditional method U-Net (i.e. results,Experimental results shows that the proposed method outperforms traditional U-Net (i.e.,1024
"nuclei the Densely for classification. Convolutional (DCRN) is used Recurrent Connected First, Network model","First, the Densely Connected Recurrent Convolutional Network (DCRN) model is used for nuclei classification.",1025
"on two We conduct experiments benchmarks, extensive large-scale i.e.","We conduct extensive experiments on two large-scale benchmarks, i.e.",1026
facial Rejecting can of significantly quality accuracy images increase any the recognition system. low,Rejecting low quality images can significantly increase the accuracy of any facial recognition system.,1027
"of the room range Methods operating contains Today, a sensors. diverse","Methods Today, the operating room contains a diverse range of sensors.",1028
"laparoscopic source such information. the stream During is an ideal of video endoscopic interventions,","During laparoscopic interventions, the endoscopic video stream is an ideal source of such information.",1029
"quantitative is challenging high to Extracting video though, information from the due its dimensionality.","Extracting quantitative information from the video is challenging though, due to its high dimensionality.",1030
streams surgery for duration of is offer uncertain. information whether Though such the estimating sufficient,Though whether such streams offer sufficient information for estimating the duration of surgery is uncertain.,1031
transfer for (ReID). an person the extremely Cross-domain (CDTL) challenging re-identification task is learning,Cross-domain transfer learning (CDTL) is an extremely challenging task for the person re-identification (ReID).,1032
"In work seek but interpretable, a this for learnable we plain, solution.","In this work we seek for a plain, interpretable, but learnable solution.",1033
variational based adjusted an inference. include stochastic variational message nonconjugate These and on pseudolikelihood passing,These include nonconjugate variational message passing based on an adjusted pseudolikelihood and stochastic variational inference.,1034
for alternatives to These approximation. attractive MCMC fast methods provide posterior,These methods provide attractive fast alternatives to MCMC for posterior approximation.,1035
"optimal technique. the using map cut obtained graph Finally, depth is","Finally, the optimal depth map is obtained using graph cut technique.",1036
advance the clinical believe brain method diagnosis. We proposed performance can the segmentation in and,We believe the proposed method can advance the performance in brain segmentation and clinical diagnosis.,1037
proposed using processes model visual information The two pathways.,The proposed model processes visual information using two pathways.,1038
subdivision costly one per avoid bit stack The iteration operations. to only requires memory,The iteration requires only one bit per subdivision to avoid costly stack memory operations.,1039
"Besides extensive code provided. analysis, an source is mathematical","Besides an extensive mathematical analysis, source code is provided.",1040
surfaces. present and We Catmull-Clark subdivision rays a intersecting method fast efficient with for,We present a fast and efficient method for intersecting rays with Catmull-Clark subdivision surfaces.,1041
animation playback. useful patch is and quick algorithm for The editing during especially turnarounds,The algorithm is especially useful for quick turnarounds during patch editing and animation playback.,1042
"difficult. CP in constructive logical notoriously with connectors information handling is However,","However, handling logical connectors with constructive information in CP is notoriously difficult.",1043
ponens And presented modus modus then based and we tollens fuzzy fuzzy on MTP.,And then we presented fuzzy modus ponens and fuzzy modus tollens based on MTP.,1044
"information propose Normalizations. leveraging Power learning this second-order a we and In paper, similarity network","In this paper, we propose a similarity learning network leveraging second-order information and Power Normalizations.",1045
rehabilitation normal life. therapy A them to robot and a return can help assisted recover,A robot assisted rehabilitation therapy can help them recover and return to a normal life.,1046
based a the An accurate identification. requires of categorical motor pattern recognition detection intention,An accurate detection of the motor intention requires a pattern recognition based categorical identification.,1047
"different a the classification we Next, of to algorithm conducted sensitivity EMG channels. analysis","Next, we conducted a sensitivity analysis of the classification algorithm to different EMG channels.",1048
"has and recent years, the approach performance. substantial In in witnessed mindset, advancements community","In recent years, the community has witnessed substantial advancements in mindset, approach and performance.",1049
accelerating be has Compressive an to MRI technique acquisition. efficient proven for (CS) Sensing,Compressive Sensing (CS) has proven to be an efficient technique for accelerating MRI acquisition.,1050
"However, are in handling these effective occlusions. long-term techniques less","However, these techniques are less effective in handling long-term occlusions.",1051
from This raw is retrieve suspected signals. allowed motion to technique of interest,This technique is allowed to retrieve suspected motion of interest from raw signals.,1052
We the evaluate to strategies values discuss experimentally. them and augment profit,We discuss strategies to augment the profit values and evaluate them experimentally.,1053
"years, In recent patch prior recurrence has for powerful a image as emerged recovery.","In recent years, patch recurrence has emerged as a powerful prior for image recovery.",1054
"models further resulting The preserve processing. enable smooth, boundaries, and are","The resulting models are smooth, preserve boundaries, and enable further processing.",1055
represents in this at one This the learning of leveraging domain. first deep attempts,This represents one of the first attempts at leveraging deep learning in this domain.,1056
a paper modalities. present utilizing for In two method representation we this self-supervised learning different,In this paper we present a self-supervised method for representation learning utilizing two different modalities.,1057
across dates. that uncertainty The function incorporates potential varies modified estimation objective break,The modified objective function incorporates estimation uncertainty that varies across potential break dates.,1058
framework. under is limit an asymptotic provided A distribution in-fill,A limit distribution is provided under an in-fill asymptotic framework.,1059
estimator. that Carlo new results suggest the simulation outperforms least-squares Monte the estimator,Monte Carlo simulation results suggest that the new estimator outperforms the least-squares estimator.,1060
"method. achieved MAS results, the state-of-the-art to performance methods superior the the experimental From DCNN","From the experimental results, the DCNN methods achieved superior performance to the state-of-the-art MAS method.",1061
"behavior far training for not a well from Unfortunately, is network networks defined inputs set.","Unfortunately, network behavior is not well defined for inputs far from a networks training set.",1062
II for StarCraft AI. We novel modular a architecture present,We present a novel modular architecture for StarCraft II AI.,1063
macros execution. scheduler by modules all centralized of order reviews decides suggested their and A,A centralized scheduler reviews macros suggested by all modules and decides their order of execution.,1064
series keeps into updater changes An of and track instantiates environment actions. macros of executable,An updater keeps track of environment changes and instantiates macros into series of executable actions.,1065
the The Attribute DAN datasets. of on a-Pascal ImageNet performance evaluated and is,The performance of DAN is evaluated on the ImageNet Attribute and a-Pascal datasets.,1066
"model comparison with state-of-the-art the methods, Experiments show results. achieves better in that proposed","Experiments show that in comparison with state-of-the-art methods, the proposed model achieves better results.",1067
"and action robustness, skeleton-based Benefiting much its has succinctness from recognition attention. attracted recently","Benefiting from its succinctness and robustness, skeleton-based action recognition has recently attracted much attention.",1068
on several outperforms methods datasets. approach state-of-the-art Our significantly large-scale other,Our approach significantly outperforms other state-of-the-art methods on several large-scale datasets.,1069
"as a filter. this, of shape the mechanical acts audio ear For the","For this, the shape of the ear acts as a mechanical audio filter.",1070
"of Neural Networks Convolutional Modern (CNNs) complex, encompassing parameters. are millions","Modern Convolutional Neural Networks (CNNs) are complex, encompassing millions of parameters.",1071
"Their storage energy particularly on deployment exerts computational, platforms. demands, embedded and","Their deployment exerts computational, storage and energy demands, particularly on embedded platforms.",1072
sparsify or accuracy. Existing maintain CNNs to approaches to require inference retraining prune,Existing approaches to prune or sparsify CNNs require retraining to maintain inference accuracy.,1073
not some retraining Such in is contexts. feasible,Such retraining is not feasible in some contexts.,1074
"sparsification of proposing explore model-independent the In paper, this we methods. by CNNs three","In this paper, we explore the sparsification of CNNs by proposing three model-independent methods.",1075
Our on-the-fly require retraining. are no methods applied and,Our methods are applied on-the-fly and require no retraining.,1076
based Often approaches workflow machine learning for basis analysis. surgical serve as,Often machine learning based approaches serve as basis for surgical workflow analysis.,1077
to only then asked data be will points. annotate these Experts,Experts will then be asked to only annotate these data points.,1078
"(DBN). CNN Recently, active of learning Deep been by to means applied has Networks Bayesian","Recently, active learning has been applied to CNN by means of Deep Bayesian Networks (DBN).",1079
uncertainties to make assign it networks These to predictions. possible,These networks make it possible to assign uncertainties to predictions.,1080
"Furthermore, analysis. workflow network a we to this recurrent extend by architecture, using surgical video-based","Furthermore, by using a recurrent architecture, we extend this network to video-based surgical workflow analysis.",1081
"progress, has folding challenges but good protein very made remain. two Contact-assisted","Contact-assisted protein folding has made very good progress, but two challenges remain.",1082
technologies. present-day of many an phase Interferometric is coherent part (InPhase) important imaging imaging,Interferometric phase (InPhase) imaging is an important part of many present-day coherent imaging technologies.,1083
size storage the clouds. the of regarding point issue Another,Another issue regarding the storage size of the point clouds.,1084
shows result RoarNet methods to Our that that publicly are state-of-the-art available. has performance superior,Our result shows that RoarNet has superior performance to state-of-the-art methods that are publicly available.,1085
available RoarNet models. implemented Tensorflow and publicly pre-trained with in is,RoarNet is implemented in Tensorflow and publicly available with pre-trained models.,1086
face Aging dramatically. recognition or variation affect the can gender performance,Aging or gender variation can affect the face recognition performance dramatically.,1087
gold has calcium coronary detection. been slice-wise the as standard manual of delineation regarded The,The slice-wise manual delineation has been regarded as the gold standard of coronary calcium detection.,1088
the Baseline assessed same performance was cohort. on,Baseline performance was assessed on the same cohort.,1089
aroused Deep have Sensing great (DNNs) Networks (CS) Compressed in attention restoration. Neural,Deep Neural Networks (DNNs) have aroused great attention in Compressed Sensing (CS) restoration.,1090
is information Human emotion forensics. the field and recognition biometric an of direction important in,Human emotion recognition is an important direction in the field of biometric and information forensics.,1091
"research on most the based However, emotion human existing single are RGB view.","However, most existing human emotion research are based on the single RGB view.",1092
"be our video-emotion first may this RGBD best the dataset. knowledge, To","To our best knowledge, this may be the first RGBD video-emotion dataset.",1093
extended our methods And can views. easily than two to more be,And our methods can be easily extended to more than two views.,1094
our state-of-art methods. some The over indicate effectiveness the results experiment methods of,The experiment results indicate the effectiveness of our methods over some state-of-art methods.,1095
"etc.) (Photoshop, the Instagram, influence conventional that tools we Apart also from of photo-editing study","Apart from that we also study the influence of conventional photo-editing tools (Photoshop, Instagram, etc.)",1096
daily a wide on audience by memorability. used,used daily by a wide audience on memorability.,1097
"directly. this of designers, and results benefit the advertisers Photographers, will from study","Photographers, designers, and advertisers will benefit from the results of this study directly.",1098
most one cancer Breast women the dangerous worldwide. and cancers is of impacting common,Breast cancer is one of the most common and dangerous cancers impacting women worldwide.,1099
"is intelligence into develop safely research artificial Accordingly, increasingly important. to methods","Accordingly, research into methods to develop artificial intelligence safely is increasingly important.",1100
"mechanisms. stovepiping the in blindspot the Additionally, a illuminate developmental for containment we potential of","Additionally, we illuminate the potential for a developmental blindspot in the stovepiping of containment mechanisms.",1101
automated We Pressure for Positive (PAP) sizing fully a present masks. nasal Airway system,We present a fully automated system for sizing nasal Positive Airway Pressure (PAP) masks.,1102
"we function an recipe, If objective model quality the the of as problems several arise.","If we model as an objective function the quality of the recipe, several problems arise.",1103
"available. gradients all model No analytical no are so recipes, expression can the","No analytical expression can model all the recipes, so no gradients are available.",1104
"words, other in it objective function noise. contains is subjective, The","The objective function is subjective, in other words, it contains noise.",1105
"Moreover, both in expensive resources. human time and are evaluations","Moreover, evaluations are expensive both in time and human resources.",1106
problems ideal as methodology an tackle (BO) with these characteristics. Bayesian emerges to Optimization,Bayesian Optimization (BO) emerges as an ideal methodology to tackle problems with these characteristics.,1107
empirical the experiments adequacy We two methodology. support of with that the evidence provide,We provide empirical evidence with two experiments that support the adequacy of the methodology.,1108
"method we Posterior-CRF. this new called training paper, a propose In end-to-end","In this paper, we propose a new end-to-end training method called Posterior-CRF.",1109
"laborious, Manual use. impractical and is segmentation for large-scale","Manual segmentation is laborious, and impractical for large-scale use.",1110
is The proposed (P-SiBA) lung from approach cases the assessed PLS-SiBA-Net lesion LesionVis. on,The proposed PLS-SiBA-Net (P-SiBA) approach is assessed on the lung lesion cases from LesionVis.,1111
"The low-compute results for such presented model and a the discussed. implementing is algorithm low-cost,","The algorithm for implementing such a low-cost, low-compute model is presented and the results discussed.",1112
"boring creative or song terribly verses), be companion, a both.","song verses), be a terribly boring creative companion, or both.",1113
hallucination Most for are face complete the designed methods of inputs.,Most of the face hallucination methods are designed for complete inputs.,1114
"by face Inspired obscured this network(OFHNet). we propose an fact, hallucination","Inspired by this fact, we propose an obscured face hallucination network(OFHNet).",1115
The low-resolution(LR) obscured images. face inpainting the network restores,The inpainting network restores the low-resolution(LR) obscured face images.,1116
the The following inpainting upsample upsampling is to network output of network.,The following upsampling network is to upsample the output of inpainting network.,1117
present gap. also allocation for approaches sample initial exploiting this We,We also present initial sample allocation approaches for exploiting this gap.,1118
frontend different be imperfections in the used them. of to can RF distinguish transmitters The,The imperfections in the RF frontend of different transmitters can be used to distinguish them.,1119
This transmitter fingerprints. identification is called RF process using,This process is called transmitter identification using RF fingerprints.,1120
"we their use this learning to characteristics. In using transmitters work, deep different nonlinear identify","In this work, we use deep learning to identify different transmitters using their nonlinear characteristics.",1121
able study variability on the of We impact to accuracy. transmitter were also identification,We were also able to study the impact of transmitter variability on identification accuracy.,1122
were compared confirmed experiments results trends. Simulation similar with which,Simulation results were compared with experiments which confirmed similar trends.,1123
detectors are State-of-the-art trackers fast. object and developing,State-of-the-art object detectors and trackers are developing fast.,1124
Trackers bear than the are more risk but efficient detectors in drifting. general of,Trackers are in general more efficient than detectors but bear the risk of drifting.,1125
"since the This baseline, on should however, detection the depend frequency quality. suboptimal tracking is","This baseline, however, is suboptimal since the detection frequency should depend on the tracking quality.",1126
"a an for empirical Finally, created optimization we automatic using method. mechanism hyperparameter","Finally, we created a mechanism for automatic hyperparameter optimization using an empirical method.",1127
first is our its technique this knowledge To of kind.,To our knowledge this technique is first of its kind.,1128
the the Activation area Further to Mapping we (CAM). attention used Class identify of,Further to identify the area of the attention we used Class Activation Mapping (CAM).,1129
state-of-the-art were for then classifiers. features training used five machine learning The selected,The selected features were then used for training five state-of-the-art machine learning classifiers.,1130
green fluoresent traced from regions using proteins. brain projections were Axonal,Axonal projections from brain regions were traced using green fluoresent proteins.,1131
resulting The data common reference were registered space. to a three-dimensional,The resulting data were registered to a common three-dimensional reference space.,1132
out the group. homology of first We generators work,We work out generators of the first homology group.,1133
"the thalamus. striatum the through and Thousands go loops the isocortex, of","Thousands of loops go through the isocortex, the striatum and the thalamus.",1134
procedure training optimization algorithm divergence and alternating contrastive is A novel integrating also an developed.,A novel training algorithm integrating contrastive divergence and an alternating optimization procedure is also developed.,1135
of as the expressive MPORBM function The investigated. is the of also power MPO-rank a,The expressive power of the MPORBM as a function of the MPO-rank is also investigated.,1136
model this work. digits for in proposed is interpretable handwritten synthesis An generative,An interpretable generative model for handwritten digits synthesis is proposed in this work.,1137
training is to process mechanism underlying complex difficult and The the is explain.,The training process is complex and the underlying mechanism is difficult to explain.,1138
"words, other a is whitening In it process.","In other words, it is a whitening process.",1139
"lower, complexity is is the Its design whole explainable. and design process significantly","Its design complexity is significantly lower, and the whole design process is explainable.",1140
data limitation. avoids effectively the images the scale and of diversity boosts It,It effectively boosts the diversity of images and avoids the data scale limitation.,1141
"on To simulate distances, controllable angles processes projection parameterized adopt transformation. different we viewing and","To simulate different viewing angles and distances, we adopt controllable parameterized projection processes on transformation.",1142
"neural call our system this We approach (re-)rendering, ""LookinGood"". and live","We call this approach neural (re-)rendering, and our live system ""LookinGood"".",1143
"a manual require First, propose we not method that ground-truth does self-supervised annotation. training","First, we propose a self-supervised training method that does not require manual ground-truth annotation.",1144
"generate temporally difference between results minimizing the two stable explicitly we frames. Finally, consecutive by","Finally, we generate temporally stable results by explicitly minimizing the difference between two consecutive frames.",1145
"and Through unseen experimentation, subjects. how generalizes extensive sequences across system demonstrate our we","Through extensive experimentation, we demonstrate how our system generalizes across unseen sequences and subjects.",1146
of Networks (GANs) field are in Adversarial generation. Generative image state-of-the-art the considered the,Generative Adversarial Networks (GANs) are considered the state-of-the-art in the field of image generation.,1147
were and models. pre-trained different Five analyzed using architectures clean CNN,Five different CNN architectures were analyzed using clean and pre-trained models.,1148
"been forms its have Depressive of sought. long and phases Biomarkers Major Disorder(MDD),","Biomarkers of Major Depressive Disorder(MDD), its phases and forms have long been sought.",1149
running EEG SampEn Artefact-free and algorithms. analyzed developed HFD by programs epochs were in-house,Artefact-free EEG epochs were analyzed by in-house developed programs running HFD and SampEn algorithms.,1150
higher complexity healthy patients to compared Depressed HFD and SampEn subjects. had,Depressed patients had higher HFD and SampEn complexity compared to healthy subjects.,1151
of in differed depressive two between parietal phases frontal complexity The and disorder. the regions,The complexity in frontal and parietal regions differed between the two phases of depressive disorder.,1152
higher than cortical areas. HFD in manifested sensitivity some SampEn,SampEn manifested higher sensitivity than HFD in some cortical areas.,1153
of EEG the distinguish between Complexity three measures groups.,Complexity measures of EEG distinguish between the three groups.,1154
article multi-processor a describes time-interval The modular in balancing scheduling composite jobs. special of,The article describes a special time-interval balancing in multi-processor scheduling of composite modular jobs.,1155
just-in-time This to planning problem approach. close scheduling is,This scheduling problem is close to just-in-time planning approach.,1156
"presented and scheduling scheduling on brief due-data/due-window surveys First, are problems. just-in-time literature","First, brief literature surveys are presented on just-in-time scheduling and due-data/due-window scheduling problems.",1157
example described. real modular world for illustrative is planning The home-building,The illustrative real world planning example for modular home-building is described.,1158
modified assembly balance to plan satisfy The to has requirements. the be,The assembly plan has to be modified to satisfy the balance requirements.,1159
implements based metaheuristic on optimization approach. The a framework local,The framework implements a metaheuristic based on local optimization approach.,1160
"information applications chain Two briefly systems) are management, transmission other described. (supply","Two other applications (supply chain management, information transmission systems) are briefly described.",1161
"two-fold: i.e. i) contribution counterfactuals, is we positive propose Our","Our contribution is two-fold: i) we propose positive counterfactuals, i.e.",1162
"domain and problems. an reimplemented architecture reinterpreted project, the present We Copycat solving letter analogy","We present the reinterpreted and reimplemented Copycat project, an architecture solving letter analogy domain problems.",1163
"Finally, of for cognitive limitations are architectures and the lines benefits analyzed along Copycat.","Finally, benefits and limitations are analyzed for cognitive architectures along the lines of Copycat.",1164
"we with (CNNs), neural scene convolutional Integrating labeling develop system. an end-to-end networks","Integrating with convolutional neural networks (CNNs), we develop an end-to-end scene labeling system.",1165
on images. aims Rain rain the to rain removal remove streaks,Rain removal aims to remove the rain streaks on rain images.,1166
mostly methods Neural based The state-of-the-art Convolutional Network~(CNN). on are,The state-of-the-art methods are mostly based on Convolutional Neural Network~(CNN).,1167
of tremendous methods made have Sustainable progress. construction,Sustainable methods of construction have made tremendous progress.,1168
"Then, background we neural of subtraction. deep the for networks discuss adequacy","Then, we discuss the adequacy of deep neural networks for background subtraction.",1169
"new to approach suggests contrast, this this a problem. In paper","In contrast, this paper suggests a new approach to this problem.",1170
"viewing Secondly, ""spherical"" sphere. we propose on the novel whole consistency constraint photometric a","Secondly, we propose a novel ""spherical"" photometric consistency constraint on the whole viewing sphere.",1171
activities. facial eyes during distinct concentrate cognitive different Human regions,Human eyes concentrate different facial regions during distinct cognitive activities.,1172
utilising to different different We into facial study classify expressions facial maps visual emotions. saliency,We study utilising facial visual saliency maps to classify different facial expressions into different emotions.,1173
frequencies is (or The aside its of) leave to theory outside blamed framework.,The theory is blamed to leave frequencies outside (or aside of) its framework.,1174
the significant security recognition is to Face anti-spoofing of face systems.,Face anti-spoofing is significant to the security of face recognition systems.,1175
anti-spoofing. effectiveness learning face Previous proved works have on supervised depth the for,Previous works on depth supervised learning have proved the effectiveness for face anti-spoofing.,1176
"as the considered they frame. the Nevertheless, an in supervision depth auxiliary only single","Nevertheless, they only considered the depth as an auxiliary supervision in the single frame.",1177
by multiple extracted feature. the backbone layers) base as,multiple layers) extracted by backbone as the base feature.,1178
"good can simple Without this additional large-scale on several method benchmarks. constraints, results achieve","Without additional constraints, this simple method can achieve good results on several large-scale benchmarks.",1179
Existing sources hop. only employ attention among one partial approaches data in interactions,Existing approaches only employ partial interactions among data sources in one attention hop.,1180
"context addition, the stage. during takes retrieval choices it into consideration answer In","In addition, it takes answer choices into consideration during the context retrieval stage.",1181
achieves accuracy state-of-the-art dataset. framework MovieQA on Our HMMN,Our HMMN framework achieves state-of-the-art accuracy on MovieQA dataset.,1182
of different Extensive ablation importance reasoning the attention holistic of contributions studies show strategies. and,Extensive ablation studies show the importance of holistic reasoning and contributions of different attention strategies.,1183
We COCO our evaluate detection PASCAL MS method VOC dataset. on benchmark and,We evaluate our method on PASCAL VOC detection benchmark and MS COCO dataset.,1184
Convolutional great tasks. have in vision success achieved various networks,Convolutional networks have achieved great success in various vision tasks.,1185
structure. research network amount of considerable to due a This on is mainly,This is mainly due to a considerable amount of research on network structure.,1186
"we on architectures, on the In study, focusing this instead focused unit convolution of itself.","In this study, instead of focusing on architectures, we focused on the convolution unit itself.",1187
"by single After generates image image, the frontal a normalizing face it input image. a","After input a single face image, it generates a frontal image by normalizing the image.",1188
method (DA) cross-domain domain a novel unsupervised for present recognition. adaptation visual We,We present a novel unsupervised domain adaptation (DA) method for cross-domain visual recognition.,1189
and medicine a relationship. intelligence longstanding Artificial have proficuous and,Artificial intelligence and medicine have a longstanding and proficuous relationship.,1190
many possibilities third-generation combine data There to and next-generation sequencing. from are,There are many possibilities to combine data from next-generation and third-generation sequencing.,1191
"finding solutions for scheduling. problems goals, with tools well-defined highly-effective Mathematical offers notably optimization for","Mathematical optimization offers highly-effective tools for finding solutions for problems with well-defined goals, notably scheduling.",1192
video saliency. show predicting results that ten in MM-Net outperforms models state-of-the-art Experimental aerial,Experimental results show that MM-Net outperforms ten state-of-the-art models in predicting aerial video saliency.,1193
transforms. This represent the using to proposes in algorithms paper votes Hough sketch,This paper proposes using sketch algorithms to represent the votes in Hough transforms.,1194
from a new to learn propose identity the We network named information set. GaitSet,We propose a new network named GaitSet to learn identity information from the set.,1195
state-of-the-art represent These recognition results accuracy. new,These results represent new state-of-the-art recognition accuracy.,1196
"robustness. of complex significant scenarios, a level model various exhibits our On","On various complex scenarios, our model exhibits a significant level of robustness.",1197
large best outperform by a These the existing margin. methods,These outperform the existing best methods by a large margin.,1198
"Moreover, data to require of models large a accurate train. amount","Moreover, accurate models require a large amount of data to train.",1199
"achieve Analysis. we To use Level Error this,","To achieve this, we use Error Level Analysis.",1200
is Convolutional at Network Neural image good classification.,Convolutional Neural Network is good at image classification.,1201
"image found quality vulnerable be to is However, it degradation. to","However, it is found to be vulnerable to image quality degradation.",1202
"is only retraining. in situation DCT-Net further applied generic trained once without and ""blindly"" The","The DCT-Net is trained ""blindly"" only once and applied in generic situation without further retraining.",1203
the in learnable Instead filters. with consider wavelet filtering we domain,Instead we consider filtering in the wavelet domain with learnable filters.,1204
established work has image-to-depth learning baselines domain. Previous in unsupervised strong the in,Previous work in unsupervised image-to-depth learning has established strong baselines in the domain.,1205
"all handle motion approaches, those approach including The proposed that outperforms state-of-the-art e.g.","The proposed approach outperforms all state-of-the-art approaches, including those that handle motion e.g.",1206
a Intel$\circledR$ available OpenVINO$^{TM}$ part of is publicly It as Toolkit.,It is publicly available as a part of Intel$\circledR$ OpenVINO$^{TM}$ Toolkit.,1207
well the intracellular that kinetics network an stochastic. known It biochemical is of is,It is well known that the kinetics of an intracellular biochemical network is stochastic.,1208
by the evolutionary illustrate model. approach ad We minimalistic the hoc,We illustrate the approach by the minimalistic ad hoc evolutionary model.,1209
in student-teacher train We that an models of a two together ensemble fashion. work,We train an ensemble of two models that work together in a student-teacher fashion.,1210
data. from YouTube) generated and downloaded artificially,downloaded from YouTube) and artificially generated data.,1211
datasets. evaluate the SC-GAN framework on proposed baseline of a set We,We evaluate the proposed SC-GAN framework on a set of baseline datasets.,1212
explanatory new using prior for Abductive reasoning observations generates knowledge. hypotheses,Abductive reasoning generates explanatory hypotheses for new observations using prior knowledge.,1213
"symbols. are contain can concept, signature specified Non-abducibles but by role, forgetting a which not","Non-abducibles are specified by a forgetting signature which can contain concept, but not role, symbols.",1214
The disjuncts. of set and are semantically hypotheses consist each of resulting minimal a,The resulting hypotheses are semantically minimal and each consist of a set of disjuncts.,1215
for elimination to Two practical and approaches approximate. redundancy full use: explored are,Two approaches to redundancy elimination are explored for practical use: full and approximate.,1216
other of for UAVs significance embedded vehicle and mobile is platforms. detection great On-board real-time,On-board real-time vehicle detection is of great significance for UAVs and other embedded mobile platforms.,1217
"obtained. progressive toric-B\'{e}zier iterative this the property the is approximation By result, curve generalized of","By this result, the progressive iterative approximation property of the generalized toric-B\'{e}zier curve is obtained.",1218
is of reduction the Dimensionality within ECG of signals considered representation. sparse framework,Dimensionality reduction of ECG signals is considered within the framework of sparse representation.,1219
approaches principled provide exploration-exploitation Reinforcement Bayesian trade-off Learning. the a in solution to,Bayesian approaches provide a principled solution to the exploration-exploitation trade-off in Reinforcement Learning.,1220
"approaches, either assume a or observable environment poorly. however, Typical fully scale","Typical approaches, however, either assume a fully observable environment or scale poorly.",1221
"we the In designing process. paper, to warehouse automate this aim layout","In this paper, we aim to automate the warehouse layout designing process.",1222
neuron a stream A impulses. Poisson with of is stimulated excitatory,A neuron is stimulated with a Poisson stream of excitatory impulses.,1223
is to stream. from the input impulse This identical those delivered,This impulse is identical to those delivered from the input stream.,1224
course of type $p(t)$ peculiarity. has a $\delta$-function The,The course of $p(t)$ has a $\delta$-function type peculiarity.,1225
encode can of neural deep layers Different networks(CNNs) different-level convolutional information.,Different layers of deep convolutional neural networks(CNNs) can encode different-level information.,1226
"High-layer features more features semantic low-layer detail contain and contain information, always information. more","High-layer features always contain more semantic information, and low-layer features contain more detail information.",1227
"from However, features background low-layer the ambiguity. semantic suffer clutter and","However, low-layer features suffer from the background clutter and semantic ambiguity.",1228
mechanism effectively The drawbacks. proposed above-mentioned the connection can overcome,The proposed connection mechanism can effectively overcome the above-mentioned drawbacks.,1229
clipping. pixel values recover network We to a to present deep lost,We present a deep network to recover pixel values lost to clipping.,1230
pixel's channels. all) of Clipping in color (or the occur any may,Clipping may occur in any (or all) of the pixel's color channels.,1231
and the of Global stability equilibria discussed. immigration-free is infection-free,Global stability of the immigration-free and infection-free equilibria is discussed.,1232
short-term numerical explored. and generalized framework is immigration scenarios are specific established A,A generalized numerical framework is established and specific short-term immigration scenarios are explored.,1233
"work, In this problem. the investigate but under-studied efficiency critical pose practically model we","In this work, we investigate the under-studied but practically critical pose model efficiency problem.",1234
"this (FPD) model we Pose To present strategy. Fast a new learning Distillation end,","To this end, we present a new Fast Pose Distillation (FPD) model learning strategy.",1235
a is effectively It the teacher by structure achieved strong network. transferring pose knowledge of,It is achieved by effectively transferring the pose structure knowledge of a strong teacher network.,1236
of network neural siamese problem. as be optimization can simple a The treated binary classification,The optimization of siamese neural network can be treated as a simple binary classification problem.,1237
"here collaboration, three are privacy. opportunities and unification The challenges summarized and concepts: in","The opportunities and challenges are here summarized in three concepts: collaboration, unification and privacy.",1238
"devices, can therefore collaboration We methods be between need allows ambiguities that which resolved. such","We therefore need methods which allows collaboration between devices, such that ambiguities can be resolved.",1239
baselines a outperform competitive report datasets. and We metrics that range of in results wide,We report results that outperform competitive baselines in a wide range of metrics and datasets.,1240
main collision Reliable requirements of for autonomous driving. is the one avoidance,Reliable collision avoidance is one of the main requirements for autonomous driving.,1241
"major a Here, every tracker. data for is association challenge multi-target","Here, data association is a major challenge for every multi-target tracker.",1242
other terms GDPF and stability. that in accuracy Real-world tests outperforms multi-target show tracker of,Real-world tests show that GDPF outperforms other multi-target tracker in terms of accuracy and stability.,1243
recognises privacy. using that a paper This activities human retains sensor,This paper recognises human activities using a sensor that retains privacy.,1244
"introduce evaluate a and the new dataset, We it. also method on","We also introduce a new dataset, and evaluate the method on it.",1245
saliency using We eye images datasets. predictions tracking NSWAM tested from several,We tested NSWAM saliency predictions using images from several eye tracking datasets.,1246
"saliency. specifically reproduce we are models outperform that designed Moreover, to other exclusively biologically-inspired saliency","Moreover, we outperform other biologically-inspired saliency models that are specifically designed to exclusively reproduce saliency.",1247
"for array propose adjustment In phase-only paper, constraint. with this response fast method we a","In this paper, we propose a fast method for array response adjustment with phase-only constraint.",1248
presented the are examples algorithm. of demonstrate effectiveness Representative proposed the to,Representative examples are presented to demonstrate the effectiveness of the proposed algorithm.,1249
torus nowadays very points on are common. a,points on a torus are nowadays very common.,1250
torus. scattered points appropriate models on are wrapped data to p-dimensional often describe Multivariate,Multivariate wrapped models are often appropriate to describe data points scattered on p-dimensional torus.,1251
These often and are solutions tasks. architecture-specific transfer not other to do,These solutions are often architecture-specific and do not transfer to other tasks.,1252
images. a This synthetic experimentation saliency model psychophysical proposes for evaluation new basis using,This experimentation proposes a new psychophysical basis for saliency model evaluation using synthetic images.,1253
control where real signals that capture data the reconstructed The evaluated wrist's glove by kinematics.,The reconstructed control signals where evaluated by real glove data that capture the wrist's kinematics.,1254
methods. matrix model slightly proposed results The than \ac{ctd} better was factorisation,The proposed \ac{ctd} model results was slightly better than matrix factorisation methods.,1255
"capability ready for applications. ""scan-and-simulate"" industrial approach makes practical, our unique scalable large-scale This and","This unique ""scan-and-simulate"" capability makes our approach scalable and practical, ready for large-scale industrial applications.",1256
inherited degenerations The of due retinal death. are photoreceptor cell to majority,The majority of inherited retinal degenerations are due to photoreceptor cell death.,1257
measured patterns. However fine-grained optogenetically-reactivated been of spatial retinas with not the has stimulation resolution,However the spatial resolution of optogenetically-reactivated retinas has not been measured with fine-grained stimulation patterns.,1258
above blindness. of is acuity the obtained limit The legal,The obtained acuity is above the limit of legal blindness.,1259
result future for is promising high studies. resolution clinical spatial This a,This high spatial resolution is a promising result for future clinical studies.,1260
"{\alpha}-stable will its the distributed under robustness, To symmetric be noise examined illustrate algorithm environment.","To illustrate its robustness, the algorithm will be examined under symmetric {\alpha}-stable distributed noise environment.",1261
"with algorithms. MUSIC-like, proposed adaptive and the compared MUSIC, is framework original The SSCM-MUSIC FLOM-MUSIC,","The proposed adaptive framework is compared with the original MUSIC-like, MUSIC, FLOM-MUSIC, and SSCM-MUSIC algorithms.",1262
Experimental improvement rotated significant a results text. show on detecting,Experimental results show a significant improvement on detecting rotated text.,1263
the encodes salient The global branch global representations.,The global branch encodes the global salient representations.,1264
scoring with superior standard methods. when Our Gleason performance compared approach achieves,Our approach achieves superior performance when compared with standard Gleason scoring methods.,1265
object an recognition life. Visual plays essential role human in daily,Visual object recognition plays an essential role in human daily life.,1266
state-of-the-art networks. for comparison deep the allows a then with neural This fair,This then allows for a fair comparison with the state-of-the-art deep neural networks.,1267
with is validated MIT-BIH proposed framework the The database. Arrhythmia,The proposed framework is validated with the MIT-BIH Arrhythmia database.,1268
stochastic without with recourse a stage We loss a single strictly function. program convex consider,We consider a single stage stochastic program without recourse with a strictly convex loss function.,1269
grid points. with and We space assume compact a of set a it decision finite,We assume a compact decision space and grid it with a finite set of points.,1270
"from datasets i.e. the two for common also We JS-MA WTCCC applied GWAS to diseases,","We also applied JS-MA to the GWAS datasets from WTCCC for two common diseases, i.e.",1271
"information demonstrated effective to computer been Moreover, be many has for vision tasks. depth","Moreover, depth information has been demonstrated to be effective for many computer vision tasks.",1272
"can flexibility, segmentation challenges methods. algorithms, particularly lead to for This however, supervised substantial","This flexibility, however, can lead to substantial challenges for segmentation algorithms, particularly supervised methods.",1273
modeled evolution by age-structured discrete time time model. population a this is stochastic The of,The time evolution of this population is modeled by a discrete time stochastic age-structured model.,1274
affects recruitment. We that incertitude only assume the,We assume that incertitude only affects the recruitment.,1275
to popularity The possible gesture applications. wide more based skeleton gaining recognition is due its,The skeleton based gesture recognition is gaining more popularity due to its wide possible applications.,1276
architecture. module a neural standard network can be learning-based included is that into This,This is a learning-based module that can be included into standard neural network architecture.,1277
The features the that abstract the actions and given. assumes formulation are,The formulation assumes that the features and the abstract actions are given.,1278
"to limitation this showing automatically. we by In address this them learn work, how","In this work, we address this limitation by showing how to learn them automatically.",1279
experimental on guarantees results domains are reported. are and several Correctness given,Correctness guarantees are given and experimental results on several domains are reported.,1280
its expression regulation nonequilibrium a and process. stochastic is Gene,Gene expression and its regulation is a nonequilibrium stochastic process.,1281
with several involved process steps molecules are copies. in in this biochemical low Different,Different molecules are involved in several biochemical steps in this process with low copies.,1282
the studies nonequilibrium also that Several cost. processes biochemical require energy show,Several studies also show that the nonequilibrium biochemical processes require energy cost.,1283
pairs. part-relational aims appearance PRN to local determine The facial feature from structure,The PRN aims to determine facial part-relational structure from local appearance feature pairs.,1284
"with appearance relational the global further improve pairwise To combined the accuracy, feature. we features","To further improve accuracy, we combined the global appearance features with the pairwise relational feature.",1285
"graph a each tracklet which treats define model First, a vertex. as we","First, we define a graph model which treats each tracklet as a vertex.",1286
"are into Afterwards, IDs. groups object tracklets the clustered which individual represent","Afterwards, the tracklets are clustered into groups which represent individual object IDs.",1287
The on excellent and resulting performance network achieves miniImageNet tieredImageNet. both,The resulting network achieves excellent performance on both miniImageNet and tieredImageNet.,1288
on introduced. are these prior implications research The inconsistencies potential of,The potential implications of these inconsistencies on prior research are introduced.,1289
"a new propose automatic Next, for scheme protocol. we evaluation subsetting","Next, we propose a new automatic subsetting scheme for evaluation protocol.",1290
results presented. MORPH-II are experimental on Preliminary,Preliminary experimental results on MORPH-II are presented.,1291
detection approach demonstrate and also of joint given removal. potential the shadow the for We,We also demonstrate the potential of the given approach for joint shadow detection and removal.,1292
"different different from different architecture. are networks results with However, very models and the system","However, the results from different networks and models are very different with different system architecture.",1293
"model with we After high product model. getting a build accuracy, a","After getting a model with high accuracy, we build a product model.",1294
extract into model human face The image will a a it vector. take and,The model will take a human face image and extract it into a vector.,1295
detection regression methods as anchor references. have Most boxes adopted current,Most current detection methods have adopted anchor boxes as regression references.,1296
"the sensitive setting detection However, of the performance boxes. to is anchor the","However, the detection performance is sensitive to the setting of the anchor boxes.",1297
detection block building Object vision. in been a computer has,Object detection has been a building block in computer vision.,1298
"images such especially aerial for from are natural of issues images, great Apart importance. pronounced","Apart from natural images, such issues are especially pronounced for aerial images of great importance.",1299
"standard left (e.g. an the established While representation ventricles, especially have cardiac ventricle,","While cardiac ventricles, especially the left ventricle, have an established standard representation (e.g.",1300
"application assumption the such semantic for methods. seem This crucial but of may trivial, is","This assumption may seem trivial, but is crucial for the application of such semantic methods.",1301
"problem. to the visual-semantic Thus, is target certain examine relationship it for important a","Thus, it is important to examine the visual-semantic relationship for a certain target problem.",1302
We highly hypotheses postulate the three verify on relationship. and consequential,We postulate and verify three highly consequential hypotheses on the relationship.,1303
pivotal plays resource networks. optimization wireless role allocation for a Effective performance in,Effective resource allocation plays a pivotal role for performance optimization in wireless networks.,1304
"mixed-integer NP-hard. programming typical are allocation problems, Unfortunately, (MINLP) resource problems are which nonlinear","Unfortunately, typical resource allocation problems are mixed-integer nonlinear programming (MINLP) problems, which are NP-hard.",1305
"on problems. to MINLP general optimize"" framework based for ""learning is solving It a","It is based on a general ""learning to optimize"" framework for solving MINLP problems.",1306
(EMT) from transition perspectives: and experimental/technological two present theoretical. epithelial-to-mesenchymal the We,We present the epithelial-to-mesenchymal transition (EMT) from two perspectives: experimental/technological and theoretical.,1307
kind modifications are hardware No of required. any,No hardware modifications of any kind are required.,1308
variability is used normal The the as score. presentation the of detection attack vectors,The variability of the normal vectors is used as the presentation attack detection score.,1309
Experiments a database of carried out approx. on were,Experiments were carried out on a database of approx.,1310
of are other available to codes source The researchers. the method made,The source codes of the method are made available to other researchers.,1311
"estimation this motion we kernel method deconvolution. for propose blind an edge-based blur paper, In","In this paper, we propose an edge-based blur kernel estimation method for blind motion deconvolution.",1312
Quantifying foundation a population wildlife of sizes estimating is management. wildlife and,Quantifying and estimating wildlife population sizes is a foundation of wildlife management.,1313
"many difficulties innate However, carnivore leading their are populations. estimating cryptic, species to in","However, many carnivore species are cryptic, leading to innate difficulties in estimating their populations.",1314
UE on output the outside position decision a or the (inside The region). is,The output is a decision on the UE position (inside or outside the region).,1315
clouds. planning the point of The parameterization on depends,The planning depends on the parameterization of point clouds.,1316
effectiveness illustrate proposed results of are Experimental given to the methods.,Experimental results are given to illustrate effectiveness of the proposed methods.,1317
"means distribution long-tail species, classes annotated a that insufficiently. observed animal are and some","animal species, a long-tail distribution means that some classes are observed and annotated insufficiently.",1318
"be Additional can prohibitively expensive, e.g. labels","Additional labels can be prohibitively expensive, e.g.",1319
experts domain involved. be to because need,because domain experts need to be involved.,1320
properties hierarchy We probabilistic such encode class a of model. a into the,We encode the properties of such a class hierarchy into a probabilistic model.,1321
"and there, we encoding loss derive a function. label novel corresponding a From","From there, we derive a novel label encoding and a corresponding loss function.",1322
effective. method suggested results and that is efficient show Both our,Both results show that our suggested method is efficient and effective.,1323
particularly interact. important is competitors not also aspect or One whether,One particularly important aspect is whether or not competitors also interact.,1324
"be to humans $\boldsymbol{\mu}$-D can their identify on Thus, walking styles. used based signatures","Thus, $\boldsymbol{\mu}$-D signatures can be used to identify humans based on their walking styles.",1325
"cross (RCS) of information contain radar moving section the Additionally, subject. signatures the the about","Additionally, the signatures contain information about the radar cross section (RCS) of the moving subject.",1326
using embedding stochastic It then in neighbor t-distributed interpreted (t-SNE). dimensions is two,It is then interpreted in two dimensions using t-distributed stochastic neighbor embedding (t-SNE).,1327
often imaging contain confounded unwanted or are partially data Biological variability.,Biological imaging data are often partially confounded or contain unwanted variability.,1328
challenging surgery face two of most variations Plastic the co-variates are disguise recognition. of and,Plastic surgery and disguise variations are two of the most challenging co-variates of face recognition.,1329
is across an people Tracking cameras multiple problem. multiple open,Tracking multiple people across multiple cameras is an open problem.,1330
"to while SCT, cater methods still ICT remains Many a challenge.","Many methods cater to SCT, while ICT still remains a challenge.",1331
is to It jointly. challenging model a problem them,It is a challenging problem to model them jointly.,1332
and/or capturing for spatial appearance short-term utilized local motion The is features. branch local,The local branch is utilized for capturing local spatial appearance and/or short-term motion features.,1333
"this filter to dynamic paper, PointConv. operation, the extend new a In convolution we named","In this paper, we extend the dynamic filter to a new convolution operation, named PointConv.",1334
build PointConv be point networks. to convolutional deep on clouds can applied,PointConv can be applied on point clouds to build deep convolutional networks.,1335
the function depends in optimized choice measure. of The the the on distance objective algorithm,The objective function optimized in the algorithm depends on the choice of the distance measure.,1336
data. contexts used several analysis of widely of distributional is in It,It is widely used in several contexts of analysis of distributional data.,1337
are critical Objects their relationships understanding. image for and contents,Objects and their relationships are critical contents for image understanding.,1338
graph image. scene a provides captures these A properties that an structured of description,A scene graph provides a structured description that captures these properties of an image.,1339
main of relationship significantly the Our part the graph method scene generation task: benefits classification.,Our method significantly benefits the main part of the scene graph generation task: relationship classification.,1340
"and age encoding techniques input are such distribution compared. label data as augmentation, Training standardization,","Training techniques such as input standardization, data augmentation, and label distribution age encoding are compared.",1341
"in-depth discussion We provide task, for this pointing future challenging analysis and research. to also","We also provide in-depth analysis and discussion for this challenging task, pointing to future research.",1342
as performance. works possible build try small to networks still as acceptable Many while have,Many works try to build networks as small as possible while still have acceptable performance.,1343
"paper This Pyramid improved MobileNet, of called version describes Mobile an Network.","This paper describes an improved version of MobileNet, called Pyramid Mobile Network.",1344
(TCM) had modes recently useful benefits. of Theory characteristic that was proposed,Theory of characteristic modes (TCM) was recently proposed that had useful benefits.,1345
also was to used isolation. enhance the TCM,TCM was also used to enhance the isolation.,1346
"different to enhance on TCM, of focus the isolation. will also This papers the approaches","This papers will also focus on the different approaches of TCM, to enhance the isolation.",1347
"Cloning paper, Deep In this approach an called (DSC). we Sensor propose","In this paper, we propose an approach called Deep Sensor Cloning (DSC).",1348
identify re-identification non-overlapping views. same across to camera   aims pedestrian Person the,  Person re-identification aims to identify the same pedestrian across non-overlapping camera views.,1349
"conceptual so, be can represented doing space By Hilbert space. using","By doing so, conceptual space can be represented using Hilbert space.",1350
high space conceptual its quantum is of This terms concept representation. dimensional in inspired,This high dimensional conceptual space is quantum inspired in terms of its concept representation.,1351
recognizes of learnt daily by activities The KID knowledge the the elderly. the model,The knowledge learnt by the KID model recognizes the daily activities of the elderly.,1352
"data based GAN discriminator learn is the real trained to distribution. a In a model,","In a GAN based model, a discriminator is trained to learn the real data distribution.",1353
training approach propose does not The conventional alter the stage. we,The approach we propose does not alter the conventional training stage.,1354
both to can conditional Our and method contribute unconditional GANs.,Our method can contribute to both conditional and unconditional GANs.,1355
"present \emph{Deep Image We retargeting. Retargeting} a image for framework content-aware (\emph{DeepIR}), coarse-to-fine","We present \emph{Deep Image Retargeting} (\emph{DeepIR}), a coarse-to-fine framework for content-aware image retargeting.",1356
explore technique discrimination with tasks We respect that to representations. simultaneous each requires visualization,We explore discrimination tasks with respect to each visualization technique that requires simultaneous representations.,1357
examples. Recent demonstrated have of networks convolutional the deep adversarial vulnerability against studies neural,Recent studies have demonstrated the vulnerability of deep convolutional neural networks against adversarial examples.,1358
for premature death. is a cause Heart major failure,Heart failure is a major cause for premature death.,1359
detection. recurrent for segregated action temporal paper assembly multiple weakly-supervised This a proposes network (STAR),This paper proposes a segregated temporal assembly recurrent (STAR) network for weakly-supervised multiple action detection.,1360
"the action we into of distribution. output corresponding recurrent network Finally, neural transform the","Finally, we transform the output of recurrent neural network into the corresponding action distribution.",1361
"requires practitioners. its However, of the expert time-consuming use participation manual","However, its manual use requires the time-consuming participation of expert practitioners.",1362
labels reading standard obtained grades. using Gold were center,Gold standard labels were obtained using reading center grades.,1363
algorithm developed thereafter A efficient fast inversion for online matrix diagonal is blockwise tracking.,A fast blockwise diagonal matrix inversion algorithm is developed thereafter for efficient online tracking.,1364
when our of proposed with state-of-the-art compared Experimental the trackers. approach demonstrate results superiority other,Experimental results demonstrate the superiority of our proposed approach when compared with other state-of-the-art trackers.,1365
distortion the of loss power Thistransformation to resolving continuity. leads due and to nonuniformspatial shape,Thistransformation leads to shape distortion due to nonuniformspatial resolving power and the loss of continuity.,1366
rep-resent Theproposed omni-directional method spherical a utilizes to polyhedron views.,Theproposed method utilizes a spherical polyhedron to rep-resent omni-directional views.,1367
canalso by methods. adopted any CNN-based method The existing proposed be,The proposed method canalso be adopted by any existing CNN-based methods.,1368
"of different have In type speciation theoretical proposed. hypothesis fact, on many been","In fact, many theoretical hypothesis on different type of speciation have been proposed.",1369
Our dynamics. population be can to also framework other type of or sexual generalised reproduction,Our framework can be generalised also to sexual reproduction or other type of population dynamics.,1370
"events to it demanding tasks. other from is intensity images Therefore, generate for","Therefore, it is demanding to generate intensity images from events for other tasks.",1371
"we the full-waveform issues, solve to To those techniques resolve employ machine-learning inversion.","To resolve those issues, we employ machine-learning techniques to solve the full-waveform inversion.",1372
show TRN that The state-of-the-art. significantly the results outperforms,The results show that TRN significantly outperforms the state-of-the-art.,1373
in in is many and Detecting shadows acoustic clinical engineering images applications. important ultrasound,Detecting acoustic shadows in ultrasound images is important in many clinical and engineering applications.,1374
able shadow-focused map. dense method is a to Our confidence generate,Our method is able to generate a dense shadow-focused confidence map.,1375
shadow network from during maps to images confidence input able directly is inference. This predict,This network is able to predict shadow confidence maps directly from input images during inference.,1376
"etc. DICE, evaluation such inter-class use and correlation as metrics We","We use evaluation metrics such as DICE, inter-class correlation and etc.",1377
the our method. to verify effectiveness of,to verify the effectiveness of our method.,1378
or of images background a methods performance depends on prior observation. long-term largely Good existing,Good performance of existing methods depends largely on prior background images or a long-term observation.,1379
proposed outperforms that Extensive our existing methods. method experiments significantly demonstrate,Extensive experiments demonstrate that our proposed method significantly outperforms existing methods.,1380
"and is based discrimination, is dual to Generality be meant to it difficulty. on","Generality is meant to be dual to discrimination, and it is based on difficulty.",1381
deep are arrays communications optical space. free-space used in for commonly Detector,Detector arrays are commonly used for free-space optical communications in deep space.,1382
"at receiver. complexity the detector a However, results computational with arrays in communication larger","However, communication with detector arrays results in a larger computational complexity at the receiver.",1383
is nontrivial an (DNNs) and of Explaining predictions networks important task. neural deep,Explaining predictions of deep neural networks (DNNs) is an important and nontrivial task.,1384
method theory feature and relies values. approximates importance recent Our on Shapley,Our method relies on recent theory and approximates Shapley feature importance values.,1385
"the implemented also We of key-point-based LGLG-WPCA, called LGLG(KP)-WPCA. version","We also implemented the key-point-based version of LGLG-WPCA, called LGLG(KP)-WPCA.",1386
of This multimodal of e.g. paper orthoimagery semantic segmentation data using addresses the task,This paper addresses the task of semantic segmentation of orthoimagery using multimodal data e.g.,1387
"infrared optical RGB, model. digital surface and","optical RGB, infrared and digital surface model.",1388
training procedure training for OrthoSeg. of propose a We supervised also,We also propose a training procedure for supervised training of OrthoSeg.,1389
framework We Single a for Detector Shot state-of-the-art propose (SSD). compressing MultiBox,We propose a framework for compressing state-of-the-art Single Shot MultiBox Detector (SSD).,1390
"Sparsity addresses following Filter framework stages: Filter the in Pruning. Induction, and compression The Selection,","The framework addresses compression in the following stages: Sparsity Induction, Filter Selection, and Filter Pruning.",1391
This than compact the most model in smaller results size architectures. existing the with,This results in the model with the size smaller than most existing compact architectures.,1392
our compare multiple the of datasets evaluate performance framework and with multiple We over methods.,We evaluate the performance of our framework with multiple datasets and compare over multiple methods.,1393
"In particular, tracking we problem video treat dynamic (i.e. a as the","In particular, we treat the video tracking problem as a dynamic (i.e.",1394
"However, and conflict other. with each iso-scallop often smoothness","However, iso-scallop and smoothness often conflict with each other.",1395
framework proposes tool for a optimization. path paper new This,This paper proposes a new framework for tool path optimization.,1396
effectiveness show proposed Experimental the given to the are results of methods.,Experimental results are given to show the effectiveness of the proposed methods.,1397
Layers refine to further stream (RAL) the Residual Attention temporal The involves temporal features.,The temporal stream further involves Residual Attention Layers (RAL) to refine the temporal features.,1398
networks. results these standard to two-population randomly-connected We first and apply,We first apply these results to standard randomly-connected and two-population networks.,1399
"demonstrate proportionally size. networks of the that we obtained their capacity the Finally, with increases","Finally, we demonstrate that the capacity of the obtained networks increases proportionally with their size.",1400
problem font pairing. introduces paper of the automatic This,This paper introduces the problem of automatic font pairing.,1401
an Font novices. important design that is pairing difficult is task for,Font pairing is an important design task that is difficult for novices.,1402
font are in three There pairing. main challenges,There are three main challenges in font pairing.,1403
"human pairing rules are and difficult conventions of experts Second, font given to formalize. by","Second, rules and conventions of font pairing given by human experts are difficult to formalize.",1404
learning asymmetric similarity propose We and dual-space (ASML). two metric algorithms: pairing font k-NN,We propose two font pairing algorithms: dual-space k-NN and asymmetric similarity metric learning (ASML).,1405
These methods large-scale learn relationships fine-grained from two automatically data.,These two methods automatically learn fine-grained relationships from large-scale data.,1406
several baseline methods also designers. from based We professional rules investigate on the,We also investigate several baseline methods based on the rules from professional designers.,1407
dataset the of and methods. effectiveness proposed Experiments studies demonstrate user and our,Experiments and user studies demonstrate the effectiveness of our proposed dataset and methods.,1408
of our were Experiments with also missing network. showed multiple performed the and robustness sequences,Experiments with multiple missing sequences were also performed and showed the robustness of our network.,1409
"both in of obvious shortcomings them have alone However, practice.","However, both of them alone have obvious shortcomings in practice.",1410
the benchmarked have Pixel-Anchor the We proposed on datasets. public,We have benchmarked the proposed Pixel-Anchor on the public datasets.,1411
major recognition driving. localization on of place A for current is focus research autonomous visual,A major focus of current research on place recognition is visual localization for autonomous driving.,1412
"Although fully idea works. has this recent explored intuitive, not been in","Although intuitive, this idea has not been fully explored in recent works.",1413
the model. supervision depth in into weak) training,weak) supervision in depth training into the model.,1414
buildings. that effectively different Experimental from our to new results showed generalizes scenes model,Experimental results showed that our model effectively generalizes to new scenes from different buildings.,1415
appear in when Problems stage. did training objects these harder become even not,Problems become even harder when these objects did not appear in training stage.,1416
maintaining description generated fluency. ensures while that the in CGO the is object,CGO ensures that the object is in the generated description while maintaining fluency.,1417
"we novel LSTMs opposite two framework achieve this, To combining a in design directions.","To achieve this, we design a novel framework combining two LSTMs in opposite directions.",1418
to CGO task objects. shows advantages of the applied when novel describing,CGO shows advantages when applied to the task of describing novel objects.,1419
show ImageNet MSCOCO experimental We results on and datasets. both,We show experimental results on both MSCOCO and ImageNet datasets.,1420
knowledge offline is level into the through estimation High target extensive learning. incorporated,High level knowledge is incorporated into the target estimation through extensive offline learning.,1421
"our accuracy. box bounding By information, achieves carefully previously unseen approach integrating target-specific","By carefully integrating target-specific information, our approach achieves previously unseen bounding box accuracy.",1422
state-of-the-art Our sets final new framework a five on benchmarks. challenging tracking,Our final tracking framework sets a new state-of-the-art on five challenging benchmarks.,1423
"on a multi-task Based we video classification. for paradigm present findings, learning these","Based on these findings, we present a multi-task learning paradigm for video classification.",1424
ability cross-dataset a in It strong generalization experiments. specifically the shows,It specifically shows a strong generalization ability in the cross-dataset experiments.,1425
"the of these to graphs connecting the axonal correspond edges areas. The fibers,","The edges of the graphs correspond to the axonal fibers, connecting these areas.",1426
case study of the the this connectomes. Exactly is the in,Exactly this is the case in the study of the connectomes.,1427
boxes non-linear can defined are as effectively. that data black often very They map,They are often defined as black boxes that can map non-linear data very effectively.,1428
CNN to tries paper This learned show an image. to at how has look,This paper tries to show how CNN has learned to look at an image.,1429
"build mental surroundings. model with, first a we To might our begin of","To begin with, we might first build a mental model of our surroundings.",1430
"remains cases in CNNs and processed in represented most understanding challenging. However, the information","However, understanding the information represented and processed in CNNs remains in most cases challenging.",1431
"extrinsic gender, (i.e. factors and face ethnicity) and morphology, age, expression","face morphology, gender, age, expression and ethnicity) and extrinsic factors (i.e.",1432
provide hope of additional interpretability for tool CNNs. an that We will methodology valuable our,We hope that our methodology will provide an additional valuable tool for interpretability of CNNs.,1433
"novel approach. Adaptive a this we parameter-free propose To end, Norm Feature","To this end, we propose a novel parameter-free Adaptive Feature Norm approach.",1434
vision computer and important one most classification Object problems. detection of is the,Object detection and classification is one of the most important computer vision problems.,1435
achieved performance experimented standard dataset accuracy. and are DDSM evaluate the acceptable To,To evaluate the performance DDSM standard dataset are experimented and achieved acceptable accuracy.,1436
"in foreground vision John find of objects ""outlierness"". using Tukey-inspired We data a measure low-level","We find foreground objects in low-level vision data using a John Tukey-inspired measure of ""outlierness"".",1437
"directions issues to research two still above be addressed. the have Nevertheless, critical","Nevertheless, the above two research directions still have critical issues to be addressed.",1438
show of the our methods. effectiveness Extensive experiments proposed,Extensive experiments show the effectiveness of our proposed methods.,1439
of implications objectives. We several fundamental biotechnological these and describe biological the result to phenomena,We describe the implications of these result to several fundamental biological phenomena and biotechnological objectives.,1440
the framework. of Experiments benchmark demonstrate four datasets AEEN-HOA on standard superiority,Experiments on four standard benchmark datasets demonstrate the superiority of AEEN-HOA framework.,1441
"Iterative RIGNet. present approach an In Recurrent this for we Gating paper, called","In this paper, we present an approach for Recurrent Iterative Gating called RIGNet.",1442
with is This a existing networks. be common powerful mechanism to with broad compatibility revealed,This is revealed to be a powerful mechanism with broad compatibility with common existing networks.,1443
standard conduct datasets. retrieval comprehensive We on image experiments,We conduct comprehensive experiments on standard image retrieval datasets.,1444
state-of-the-art method performance Our weakly achieves most datasets. ASDA supervised on,Our weakly supervised ASDA method achieves state-of-the-art performance on most datasets.,1445
the for demonstrate that ASDA results effective method image The proposed retrieval. is,The results demonstrate that the proposed ASDA method is effective for image retrieval.,1446
"we paper, person the identification. iQIYI-VID, for this multi-modal In video largest introduce dataset","In this paper, we introduce iQIYI-VID, the largest video dataset for multi-modal person identification.",1447
person identification promote We have the released online to research. multi-modal dataset,We have released the dataset online to promote multi-modal person identification research.,1448
to segmentation is method powerful a understanding. facilitate visual scene Semantic,Semantic segmentation is a powerful method to facilitate visual scene understanding.,1449
means large very useful becomes scale This a to as summarize overhead imagery.,This becomes very useful as a means to summarize large scale overhead imagery.,1450
we to work In paper overhead segmentation imagery. applications this present our on semantic with,In this paper we present our work on semantic segmentation with applications to overhead imagery.,1451
compare with state-of-the-art We other datasets performance architectures. results several on report and,We report results on several datasets and compare performance with other state-of-the-art architectures.,1452
"addition, assessments on If, increases the multi-modal imposed burden substantially. in imaging considered, disease is","If, in addition, multi-modal imaging is considered, the burden imposed on disease assessments increases substantially.",1453
"integrate multi-modal almost leverage how unexplored. remains First, effectively data fully to and","First, how to effectively integrate and fully leverage multi-modal data remains almost unexplored.",1454
"insufficient. instances cognitive tasks, recognizing on individual these better merely is To performance object achieve","To achieve better performance on these cognitive tasks, merely recognizing individual object instances is insufficient.",1455
"strategy. present Learning In we (SRDL) work, a Deep this Self-Referenced","In this work, we present a Self-Referenced Deep Learning (SRDL) strategy.",1456
spatiotemporal essential Capturing recognition. topic video an dynamics in is,Capturing spatiotemporal dynamics is an essential topic in video recognition.,1457
"as motion. to capture Higher-order easier such make it context-sensitive patterns, operations","Higher-order operations make it easier to capture context-sensitive patterns, such as motion.",1458
be more operations can and generic Learnable flexible. higher-order,Learnable higher-order operations can be more generic and flexible.,1459
successes recent of generative The algorithm networks. the by idea is adversarial inspired of,The idea of the algorithm is inspired by recent successes of generative adversarial networks.,1460
and memory at computational cost high domains of requirements. the multiple across,across multiple domains at the cost of high computational and memory requirements.,1461
"filter-level CNNs. In pruning this work, present of we approach stability-based a for","In this work, we present a stability-based approach for filter-level pruning of CNNs.",1462
money the in rate reaction empirical models with interest results support function. The,The empirical results support models with money in the interest rate reaction function.,1463
These in the of degrade practical challenges applications. performance the methods existing,These challenges degrade the performance of the existing methods in practical applications.,1464
"flow this based address paper, In to framework optical this is proposed problem. an","In this paper, an optical flow based framework is proposed to address this problem.",1465
"we reconstruction. paper, live surface new collaborative In this system present a for dense","In this paper, we present a new system for live collaborative dense surface reconstruction.",1466
"the For example, of question the answer grass?"" their is ""What color to the","For example, their answer to the question ""What is the color of the grass?""",1467
"book?"" question ""Green"", is a of title like ""What usually the whereas the is","is usually ""Green"", whereas a question like ""What is the title of the book?""",1468
be biases. by statistical answered inferring cannot,cannot be answered by inferring statistical biases.,1469
VQA behaviors by learned examples also of results show models unusual in attempting tasks. Our,Our results also show examples of unusual behaviors learned by models in attempting VQA tasks.,1470
Interaction interact with important to humans understand objects. (HOI) Human-Object an is how problem Detection,Human-Object Interaction (HOI) Detection is an important problem to understand how humans interact with objects.,1471
extensively We V-COCO the evaluate datasets. method HICO-DET proposed and on,We extensively evaluate the proposed method on HICO-DET and V-COCO datasets.,1472
devices rapidly. demand on semantic model of segmentation The has mobile been increasing applying,The demand of applying semantic segmentation model on mobile devices has been increasing rapidly.,1473
CGNet and also memory of parameters to save designed the number is elaborately reduce footprint.,CGNet is also elaborately designed to reduce the number of parameters and save memory footprint.,1474
"the Under an significantly proposed equivalent of parameters, CGNet existing segmentation number networks. outperforms","Under an equivalent number of parameters, the proposed CGNet significantly outperforms existing segmentation networks.",1475
and effectiveness on the proposed approach. verify experiments of Extensive Cityscapes CamVid datasets the,Extensive experiments on Cityscapes and CamVid datasets verify the effectiveness of the proposed approach.,1476
"DNNs datasets of limit Small and performance. limit the depth implementable,","Small datasets limit the depth of DNNs implementable, and limit performance.",1477
of the scheme. analysis effectiveness Experimental and proposed results the demonstrate,Experimental results and analysis demonstrate the effectiveness of the proposed scheme.,1478
"data, An during test as training. served dataset withheld internal","An internal dataset served as test data, withheld during training.",1479
ROC ROC analysis was readout by The the evaluated = AUC). under curve (Area,The readout was evaluated by ROC analysis (Area under the ROC curve = AUC).,1480
performance by strong Retina two-staged yields counterparts. detection its reached U-Net complex only more,Retina U-Net yields strong detection performance only reached by its more complex two-staged counterparts.,1481
rates exchange eight dollar. set a the US We to apply against of model this,We apply this model to a set of eight exchange rates against the US dollar.,1482
Experiments suppress backgrounds more demonstrate alarms that caused PAN false effectively. text-like by can,Experiments demonstrate that PAN can suppress false alarms caused by text-like backgrounds more effectively.,1483
require applications try-on clothes Most images. virtual existing clean,Most existing virtual try-on applications require clean clothes images.,1484
the for we attempt this present In alleviating re-training. at need paper a first,In this paper we present a first attempt at alleviating the need for re-training.,1485
approach We different such of via show benefits several the an applications.,We show the benefits of such an approach via several different applications.,1486
"have detect surveillance behaviour several unusual to Recently, been videos. in explored techniques","Recently, several techniques have been explored to detect unusual behaviour in surveillance videos.",1487
learned the is activation The from by data dependency networks.,The dependency is learned from data by the activation networks.,1488
"applications However, pretraining for or require sophisticated training, existing sets data often large schemes.","However, existing applications often require large data sets for training, or sophisticated pretraining schemes.",1489
since integral Jewelry human part culture been an has of ages.,Jewelry has been an integral part of human culture since ages.,1490
shows appealing. generated and designs visually by are our likeable evaluation approach highly that Our,Our evaluation shows that designs generated by our approach are highly likeable and visually appealing.,1491
"this. we heuristic Following this, for two addressing review methods current","Following this, we review two current heuristic methods for addressing this.",1492
"is regions in such always However, semantic not straightforward and define it locate to videos.","However, it is not always straightforward to define and locate such semantic regions in videos.",1493
on clearly that SPCNET datasets start-of-the-art standard Experiments outperforms demonstrate our methods.,Experiments on standard datasets demonstrate that our SPCNET clearly outperforms start-of-the-art methods.,1494
important numerous in applications detection is and with human-computer an interaction. robotics Fine-grained action task,Fine-grained action detection is an important task with numerous applications in robotics and human-computer interaction.,1495
temporal model of independent jointly spatio-temporal Our (instead learns spatial using and streams). features,Our model jointly learns spatio-temporal features (instead of using independent spatial and temporal streams).,1496
"instead the from temporal The pixel feature space e.g. of is space, learned component","The temporal component is learned from the feature space instead of pixel space, e.g.",1497
"long-temporal features networks, used can flexibly e.g. with in produced other modeling be The conjunction","The produced features can be flexibly used in conjunction with other long-temporal modeling networks, e.g.",1498
"compute propose paper, a new this (GFT). we Graph regression-based Transform to Fourier In algorithm","In this paper, we propose a new regression-based algorithm to compute Graph Fourier Transform (GFT).",1499
"resulting the the desired. are always as sparse domain However, as reconstructions edge not in","However, the resulting reconstructions are not always as sparse in the edge domain as desired.",1500
"cost in provide some Iteratively extended but improvement of reweighted methods the accuracy, runtime. at","Iteratively reweighted methods provide some improvement in accuracy, but at the cost of extended runtime.",1501
for the We our provide theoretical analysis rotation-invariance by methods. also achieved,We also provide theoretical analysis for the rotation-invariance achieved by our methods.,1502
the state-of-the-art quantitative that demonstrate experiments other outperforms Both methods. proposed and method qualitative,Both qualitative and quantitative experiments demonstrate that the proposed method outperforms other state-of-the-art methods.,1503
"our describe In in this and to chest pneumonia approach we localization classification radiographs. work,","In this work, we describe our approach to pneumonia classification and localization in chest radiographs.",1504
"of Moreover, up will are memory lots resources. networks which intensive neural take hardware","Moreover, neural networks are memory intensive which will take up lots of hardware resources.",1505
feature convolutional of a find redundancy there maps from is We lot that in layers.,We find that there is a lot of redundancy in feature maps from convolutional layers.,1506
MDNet in also The the of problem time major efficiency. lies,The major problem of MDNet also lies in the time efficiency.,1507
bilinear Because can coordinates interpolation. RoIAlign by float process number,Because RoIAlign can process float number coordinates by bilinear interpolation.,1508
performed checking techniques. using is task usually model This,This task is usually performed using model checking techniques.,1509
"we verification. planning In model of paper, discuss this downside unconstrained the domain","In this paper, we discuss the downside of unconstrained planning domain model verification.",1510
convolutional objective success neural learning The essential formulation for of is networks. the,The objective learning formulation is essential for the success of convolutional neural networks.,1511
supervised through is A optimal the solution for learning. determined closed-form of the parameters integration,A closed-form solution of the optimal parameters for the integration is determined through supervised learning.,1512
"each the we before to At propose integration. individual the level, local promote diffusion","At the local level, we propose to promote each individual diffusion before the integration.",1513
relationship mathematical close the clustering. analysis diffusion spectral saliency Our and between reveals,Our mathematical analysis reveals the close relationship between saliency diffusion and spectral clustering.,1514
of necessary accurate limb is estimate for position An movement.,An accurate estimate of limb position is necessary for movement.,1515
features assumption that contain method adjacent is Our the redundant on information. based,Our method is based on the assumption that adjacent features contain redundant information.,1516
"Thus, information. can the compact spatial more use effectively is model and proposed content and","Thus, the proposed model is more compact and can effectively use spatial and content information.",1517
personality affective computing. emotion and are both to central analysis Apparent,Apparent personality and emotion analysis are both central to affective computing.,1518
"and Siamese-like an introduce To network. end-to-end PersEmoN, we end, this trainable deep","To this end, we introduce PersEmoN, an end-to-end trainable and deep Siamese-like network.",1519
dedicated personality to and their networks own are annotated Emotion dataset.,Emotion and personality networks are dedicated to their own annotated dataset.,1520
"to is sources. an adversarial-like representation heterogeneous promote among Furthermore, coherence function loss employed dataset","Furthermore, an adversarial-like loss function is employed to promote representation coherence among heterogeneous dataset sources.",1521
"explore Based emotion-to-apparent-personality on relationship. also this, the we","Based on this, we also explore the emotion-to-apparent-personality relationship.",1522
demonstrate experiments Extensive the effectiveness of PersEmoN.,Extensive experiments demonstrate the effectiveness of PersEmoN.,1523
"perceptual However, lack of there still style. artistic of components the is analysis computational of","However, there is still lack of computational analysis of perceptual components of the artistic style.",1524
several style brings of Such advantages. decomposition,Such decomposition of style brings several advantages.,1525
is a topic recently. inference of Convolutional Neural Efficient thriving Networks,Efficient inference of Convolutional Neural Networks is a thriving topic recently.,1526
this method efficient to presents address This paper an challenge.,This paper presents an efficient method to address this challenge.,1527
the aims to dehazing hazy Image a content recover image. from uncorrupted,Image dehazing aims to recover the uncorrupted content from a hazy image.,1528
"additional data. coincides stable to fails explain the with solution One but Inmarsat-result,","One stable solution coincides with the Inmarsat-result, but fails to explain additional data.",1529
"detect occluded blurry or it detectors challenging However, faces. remains tiny, for to face","However, it remains challenging for face detectors to detect tiny, occluded or blurry faces.",1530
achieved our results. model on and state-of-art We WIDER FACE dataset evaluate,We evaluate our model on WIDER FACE dataset and achieved state-of-art results.,1531
"commonalities multimodal modelling Typically, of explores structural imaging functional predictive connectomes across connectomes from data.","Typically, predictive modelling of functional connectomes from structural connectomes explores commonalities across multimodal imaging data.",1532
weighted our based diffusion MRI fMRI We validate and on data. approach resting-state,We validate our approach based on resting-state fMRI and diffusion weighted MRI data.,1533
videos. the language-based We untrimmed temporal in of address problem localization,We address the problem of language-based temporal localization in untrimmed videos.,1534
"as windows Besides, ACL also sliding capability results. has the to localization regress","Besides, ACL also has the capability to regress sliding windows as localization results.",1535
regions of over folds non-invertibility the surface itself. where,regions of non-invertibility where the surface folds over itself.,1536
for Codes this publication. paper upon be released will,Codes for this paper will be released upon publication.,1537
in couple has the seen tremendous of over past growth recognition face decades. Research,Research in face recognition has seen tremendous growth over the past couple of decades.,1538
"network, physically-based of dataset a train generate To synthetic we images. novel the","To train the network, we generate a novel dataset of physically-based synthetic images.",1539
using problem two-level screening We addressed system. a have this,We have addressed this problem using a two-level screening system.,1540
in using is usually It problem computer networks. important an neural solved vision and is,It is an important problem in computer vision and is usually solved using neural networks.,1541
"we Refinement To Network architecture. the introduce end, this Double","To this end, we introduce the Double Refinement Network architecture.",1542
attacker). target parameters architecture are the to model's when not known and,when the target model's architecture and parameters are not known to attacker).,1543
algorithms Current the level enhance seek decision on that i.e. to transferability adversarial work attack,Current attack algorithms that seek to enhance adversarial transferability work on the decision level i.e.,1544
generate decisions. the alter that perturbations network,generate perturbations that alter the network decisions.,1545
for task. cross-entropy beyond generalize original not its object softmax recognition) therefore does and,softmax cross-entropy for object recognition) and therefore does not generalize beyond its original task.,1546
"significantly This builds the ensembling a which system, the model. robustness tight of improves","This builds a tight ensembling system, which significantly improves the robustness of the model.",1547
"computational scales general, cost the GAN the of with training exponentially resolution. In","In general, the computational cost of training GAN scales exponentially with the resolution.",1548
specific train own discriminator. We sub-generator each its with,We train each sub-generator with its own specific discriminator.,1549
the low-resolution a input. problem from high-resolution We image a address blurry of face restoring,We address the problem of restoring a high-resolution face image from a blurry low-resolution input.,1550
deblurring This and as difficult super-resolution be tackled need to problem is simultaneously.,This problem is difficult as super-resolution and deblurring need to be tackled simultaneously.,1551
"the well. details CNN handle image cannot based However, method","However, the CNN based method cannot handle image details well.",1552
novel facial via further exemplar-based enhancement detail matching. We component develop algorithm a,We further develop a novel exemplar-based detail enhancement algorithm via facial component matching.,1553
show outperforms and Extensive experiments state-of-the-art qualitatively. proposed the quantitatively the method algorithms that both,Extensive experiments show that the proposed method outperforms the state-of-the-art algorithms both quantitatively and qualitatively.,1554
models detection of data. object Conventional require amounts training large,Conventional object detection models require large amounts of training data.,1555
"merely previously unseen by their objects semantic In knowing can description. humans recognize comparison,","In comparison, humans can recognize previously unseen objects by merely knowing their semantic description.",1556
"understanding significantly and Particularly, object improves it recognition. scene","Particularly, it significantly improves scene understanding and object recognition.",1557
Markov move-making algorithms (MRFs). consider We multi-label Fields of energy for Random minimization,We consider move-making algorithms for energy minimization of multi-label Markov Random Fields (MRFs).,1558
"Such include methods range-moves. {\alpha}{\beta}-swap, {\alpha}-expansion, and","Such methods include {\alpha}-expansion, {\alpha}{\beta}-swap, and range-moves.",1559
with algorithms. respect substantial previous improvement Experiments to move-making show,Experiments show substantial improvement with respect to previous move-making algorithms.,1560
"clear, and partial is into cloudy used days. index days cloudy Clearness to classify","Clearness index is used to classify days into clear, cloudy and partial cloudy days.",1561
codes and decoding decoded typically hard Staircase bounded-distance (BDD) are (SCCs) using decisions. iterative,Staircase codes (SCCs) are typically decoded using iterative bounded-distance decoding (BDD) and hard decisions.,1562
method using rigorously validated This embedding. complete is pathological,This method is rigorously validated using complete pathological embedding.,1563
was tumor evaluated percentage change against semi-automatic of volume The segmentations.,The percentage of tumor volume change was evaluated against semi-automatic segmentations.,1564
"labeled the to Readily has the synthetic potential data available, accurately effort. reduce","Readily available, accurately labeled synthetic data has the potential to reduce the effort.",1565
use occlusion problem. gain We on masking the probabilistic insight to,We use probabilistic occlusion masking to gain insight on the problem.,1566
"important estimation is blind Therefore, of for image the deblurring. the essentially blur kernel","Therefore, the estimation of the blur kernel is essentially important for blind image deblurring.",1567
"this and the magnitude, pattern the paper.) motion direction it we motion in call","the motion direction and magnitude, we call it the motion pattern in this paper.)",1568
"the kernel blur, to estimation that caused approach. new blur a efficient leading and","that caused the blur, leading to a new and efficient blur kernel estimation approach.",1569
"further blur our We which non-uniform approach blur, spatially varying extend involves to handle kernels.","We further extend our approach to handle non-uniform blur, which involves spatially varying blur kernels.",1570
"an area with applications. active research, generation possible text-to-image many of is Conditional","Conditional text-to-image generation is an active area of research, with many possible applications.",1571
approach interactive our is important toward an believe We step generation.,We believe our approach is an important step toward interactive generation.,1572
videos. frames low to video resolutions information degrade downsample surveillance We extremely private from to,We downsample video frames to extremely low resolutions to degrade private information from surveillance videos.,1573
"requiring restrictive suboptimal. However, both image-level discrete labels and is","However, requiring discrete image-level labels is both restrictive and suboptimal.",1574
"of consists Real-world usually text, as ""supervision"" unstructured captions. more such","Real-world ""supervision"" usually consists of more unstructured text, such as captions.",1575
between we work association learn images captions. maps this In and,In this work we learn association maps between images and captions.,1576
"captions detection Further, show on we results encouraging image-level only. object from","Further, we show encouraging results on object detection from image-level captions only.",1577
are by How these al. computed and is responses much-debated et beliefs represented (Fiser neural,How these beliefs are computed and represented by neural responses is much-debated (Fiser et al.,1578
CNNs combines Our to BSIF of method image. lightweight classify input views multiple the,Our method combines lightweight CNNs to classify multiple BSIF views of the input image.,1579
"agents. is the as environment channel generic the modelled Firstly, communication between","Firstly, the environment is modelled as the generic communication channel between agents.",1580
"the Secondly, agent architecture, UIC a adopts e.g. for learning-agent several unifying well-adopted model model","Secondly, the UIC model adopts a learning-agent model for unifying several well-adopted agent architecture, e.g.",1581
and explicitly transfer. implicit information transfer information,explicitly information transfer and implicit information transfer.,1582
"time, same expansion horizontal the network At (e.g.","At the same time, horizontal expansion network (e.g.",1583
network Model) Inception improve been fully has performance investigated. to not as an way alternative,Inception Model) as an alternative way to improve network performance has not been fully investigated.,1584
"of performance. leaf achieved even less nodes, all By maps features combining better feature","By combining all features of leaf nodes, even less feature maps achieved better performance.",1585
"generate the always the To of ATP needed. oscillation, hydrolysis is","To generate the oscillation, the hydrolysis of ATP is always needed.",1586
also directly sampling. extended This be to method may complementary,This method may be also directly extended to complementary sampling.,1587
of non-negative sampling only modified The patterns. consists,The modified sampling consists only of non-negative patterns.,1588
images challenge. a Detecting has significant become emerging manipulated,Detecting manipulated images has become a significant emerging challenge.,1589
"framework, design correlation independent tracking multi-hierarchical a this paper, (MHIT). called In we novel filters","In this paper, we design a novel tracking framework, called multi-hierarchical independent correlation filters (MHIT).",1590
have the any detection accuracy loss. improvements slightly of without speed increased These,These improvements have slightly increased the speed of detection without any accuracy loss.,1591
applications. improve results all Push-pull inhibition to helped considerably in,Push-pull inhibition helped to improve results considerably in all applications.,1592
"Currently, based on bag-of-words are approach. most solutions the","Currently, most solutions are based on the bag-of-words approach.",1593
applicability the by deep we paper neural this both addressing networks demonstrate In issues. of,In this paper we demonstrate the applicability of deep neural networks by addressing both issues.,1594
vision. modeling problem in challenging a is videos in fundamental yet Temporal computer,Temporal modeling in videos is a fundamental yet challenging problem in computer vision.,1595
saliency geometric computed the The map vertex is curvature. based on,The geometric saliency map is computed based on the vertex curvature.,1596
models discriminated. are and non-face face The then,The face and non-face models are then discriminated.,1597
of comparisons the The proposed experiments prove the framework. and effectiveness,The experiments and comparisons prove the effectiveness of the proposed framework.,1598
LiDAR multiple clouds realistic demonstrate practicality on our point from the We surveys. method of,We demonstrate the practicality of our method on realistic point clouds from multiple LiDAR surveys.,1599
long-length existing methods. the gives summarization for results CSNet to compared videos better Our,Our CSNet gives better summarization results for long-length videos compared to the existing methods.,1600
"in the In videos. to attention introduce addition, information handle mechanism dynamic we an","In addition, we introduce an attention mechanism to handle the dynamic information in videos.",1601
tomography unknown and establishes a This connection between (uDGP). geometry problem view unassigned distance,This establishes a connection between unknown view tomography and unassigned distance geometry problem (uDGP).,1602
background aims segments objects. foreground to BGS $\mathcal{B}$ and models segmenter construct The,The BGS segmenter $\mathcal{B}$ aims to construct background models and segments foreground objects.,1603
and on threads. $\mathcal{S}$ two in work parallel $\mathcal{B}$,$\mathcal{B}$ and $\mathcal{S}$ work in parallel on two threads.,1604
"At time, the segmenter real-time semantic ${S}_t$. $\mathcal{S}$ semantics extracts same the object-level the","At the same time, the real-time semantic segmenter $\mathcal{S}$ extracts the object-level semantics ${S}_t$.",1605
"the update background is Finally, the to model. fed FG/BG back ${D}_t$ refined mask","Finally, the refined FG/BG mask ${D}_t$ is fed back to update the background model.",1606
"very proposed flexible addition, the potential and has our for is In framework generalization.","In addition, our proposed framework is very flexible and has the potential for generalization.",1607
generator. interpretable propose We efficient an graph and scene,We propose an efficient and interpretable scene graph generator.,1608
details accuracy illustrated. the been and have various measures of Comprehensive metrics,Comprehensive details of the various accuracy measures and metrics have been illustrated.,1609
"of datasets the overview, giving For are clear purpose the a relevant discussed. characteristics of","For the purpose of giving a clear overview, the characteristics of relevant datasets are discussed.",1610
pairs. generate to realistic image introduced exposure short-long is novel A approach,A novel approach is introduced to generate realistic short-long exposure image pairs.,1611
of also fusion blur. in enables It exposure the presence motion,It also enables exposure fusion in the presence of motion blur.,1612
"lead However, misdiagnosis. to may suffers artifacts patient from caused by movements which it motion","However, it suffers from motion artifacts caused by patient movements which may lead to misdiagnosis.",1613
in problem. results optimization non-convex a This challenging,This results in a challenging non-convex optimization problem.,1614
Our and applications. contribution extensions future promises,Our contribution promises future extensions and applications.,1615
the pixels among and relation RETRaIN the active neighboring regions. encodes facial The of reference,The RETRaIN encodes the relation among the reference and neighboring pixels of facial active regions.,1616
The superior in accuracy state-of-the-art comparison achieves recognition RETRaIN to techniques.,The RETRaIN achieves superior recognition accuracy in comparison to state-of-the-art techniques.,1617
"automated This precise and imply adaptive methods using software. sensors demand and objective,","This demand imply objective, precise and automated methods using sensors and adaptive software.",1618
overcome to primary functional Compressed solutions two split CPRI this. are and,Compressed CPRI and functional split are two primary solutions to overcome this.,1619
"paper, for (VRF) introduce In this the C-RAN. of Rate Variable we concept Fronthaul","In this paper, we introduce the concept of Variable Rate Fronthaul (VRF) for C-RAN.",1620
transport C-RAN efficient a allows shared This cells of data backhaul. for more over a,This allows for a more efficient transport of C-RAN cells data over a shared backhaul.,1621
"of most only single training works type of data. exploit the However,","However, most of the works only exploit single type of training data.",1622
We that increase also significantly transfer classification can learning performance. show the,We also show that transfer learning can significantly increase the classification performance.,1623
for intuitively planning. plans and allows users more criticize to visualize mixed-initiative TGE-viz,TGE-viz allows users to visualize and criticize plans more intuitively for mixed-initiative planning.,1624
segmenting Processing usually and nuclei cytology overlapping involve cells. images,Processing cytology images usually involve segmenting nuclei and overlapping cells.,1625
"have or Recently, into captions, sentiment the as non-factual such aspects style. incorporated some models","Recently, some models have incorporated non-factual aspects into the captions, such as sentiment or style.",1626
"to better attention-based To add image to design we model sentiment captions. address this, an","To address this, we design an attention-based model to better add sentiment to image captions.",1627
with evaluation using The the in captioning sentiment metrics. standard work state-of-the-art model outperforms image,The model outperforms the state-of-the-art work in image captioning with sentiment using standard evaluation metrics.,1628
an FANet end-to-end in manner. trained The whole is,The whole FANet is trained in an end-to-end manner.,1629
"have gained including CBIR. learning vision, computer attention in methods Recently, deep","Recently, deep learning methods have gained attention in computer vision, including CBIR.",1630
for paper segmentation. an and presents jointly semantic registration facade approach solving This efficient,This paper presents an efficient approach for solving jointly facade registration and semantic segmentation.,1631
the to we registration. aim the through improve of the segmentation quality semantic Simultaneously,Simultaneously we aim to improve the quality of the semantic segmentation through the registration.,1632
two in jointly a framework. solved are Expectation-Maximization problems These,These two problems are jointly solved in a Expectation-Maximization framework.,1633
observed videos. aims human partially recognizing This at paper in actions,This paper aims at recognizing partially observed human actions in videos.,1634
"observed. contain often actions partially uncontrolled make Action environments videos frames, which in corrupt acquired","Action videos acquired in uncontrolled environments often contain corrupt frames, which make actions partially observed.",1635
"arbitrary appear of irregularly. frames last these for Furthermore, time can and lengths","Furthermore, these frames can last for arbitrary lengths of time and appear irregularly.",1636
training and performance data the action of are recognition pre-trained systems. degrade They inconsistent with,They are inconsistent with training data and degrade the performance of pre-trained action recognition systems.,1637
present this We an issue. to address approach,We present an approach to address this issue.,1638
alternatives each from data. We augment extra segment retrieved training with,We augment each segment with extra alternatives retrieved from training data.,1639
road successive two scene of consists This construction. detection and stages:,This consists of two successive stages: road detection and scene construction.,1640
by detection new Markov a is realized The field random (MRF) algorithm. superpixel road,The road detection is realized by a new superpixel Markov random field (MRF) algorithm.,1641
interaction spatio-temporally the on established superpixels. the smoothness The is basis of term adjacent of,The smoothness term is established on the basis of the interaction of spatio-temporally adjacent superpixels.,1642
"the scene construction, and are background independently. the subsequent modeled regions foreground In","In the subsequent scene construction, the foreground and background regions are modeled independently.",1643
TempCNNs the is for evaluate quantitatively and contribution goal The classification. SITS to qualitatively of,The goal is to quantitatively and qualitatively evaluate the contribution of TempCNNs for SITS classification.,1644
"e.g. in also We highlight obtained computer differences some vision, results with","We also highlight some differences with results obtained in computer vision, e.g.",1645
"produced TempCNNs. quality the the visual land cover of assess maps Finally, by we","Finally, we assess the visual quality of the land cover maps produced by TempCNNs.",1646
demonstrate performance We achieves datasets. on that challenging MonoGRNet state-of-the-art,We demonstrate that MonoGRNet achieves state-of-the-art performance on challenging datasets.,1647
an important pose task. from Hand a challenging image estimation but RGB is monocular,Hand pose estimation from a monocular RGB image is an important but challenging task.,1648
"and CMU-PS hand benchmarks, three TAGAN evaluate pose RHP, on the We STB, including datasets.","We evaluate TAGAN on three benchmarks, including the RHP, STB, and CMU-PS hand pose datasets.",1649
task many it a image is affected since factors. by challenging very Human generation is,Human image generation is a very challenging task since it is affected by many factors.,1650
on indicate Extensive experiments the our of method. and potential LIP COCO datasets,Extensive experiments on COCO and LIP datasets indicate the potential of our method.,1651
"to limitations. In this a we network paper, overcome deep novel neural propose such structure-aware","In this paper, we propose a novel structure-aware deep neural network to overcome such limitations.",1652
datasets on method. effectiveness tracking our validated proposed the of experiments Extensive five benchmark,Extensive experiments on five tracking benchmark datasets validated the effectiveness of our proposed method.,1653
is reconstruction anomaly error as indicator. used an frame The,The frame reconstruction error is used as an anomaly indicator.,1654
shown Our in effective to in be detecting videos. is model anomalies,Our model is shown to be effective in detecting anomalies in videos.,1655
approachcould in requiring a be common component sense. This applications visual,This approachcould be a component in applications requiring visual common sense.,1656
"extensively studied hashing been methods have retrieval. Recently, for large-scale similarity-preserving image","Recently, similarity-preserving hashing methods have been extensively studied for large-scale image retrieval.",1657
an identity-preserving introduce different to subjects. between that aids distinguish the discriminator loss We adversarial,We introduce an identity-preserving adversarial loss that aids the discriminator to distinguish between different subjects.,1658
centerline regularization. vessel combines constraints divergence with robust reconstruction curvature Our tree,Our vessel tree centerline reconstruction combines divergence constraints with robust curvature regularization.,1659
and spotting competitive reading The end-to-end text achieve developed also performance. systems scene very,The developed end-to-end systems also achieve very competitive scene text spotting and reading performance.,1660
It be verified resulting can that a finite estimator variance. has the,It can be verified that the resulting estimator has a finite variance.,1661
"further unified is Under this function formulation, each encode defined to objective a cue.","Under this formulation, a unified objective function is further defined to encode each cue.",1662
every system signals. physical processing which on signal in take nearly digital operates part Quantizers,Quantizers take part in nearly every digital signal processing system which operates on physical signals.,1663
then networks. estimation our in channel apply We massive technique MIMO proposed to,We then apply our proposed technique to channel estimation in massive MIMO networks.,1664
its instead of hide to when agent intentions. trying the revealing is,when the agent is trying to hide instead of revealing its intentions.,1665
is This rigorous using formulation. yet achieved probabilistic computationally-efficient a,This is achieved using a rigorous yet computationally-efficient probabilistic formulation.,1666
estimation as Point-set likelihood EM and is cast maximum a the using registration solved algorithm.,Point-set registration is cast as a maximum likelihood estimation and solved using the EM algorithm.,1667
sideslip framework novel of presents angle. paper the estimation a for This,This paper presents a novel framework for estimation of the sideslip angle.,1668
observer. proposed with conjunction in utilizes a an adaptive algorithm model model-based The tire,The proposed algorithm utilizes an adaptive tire model in conjunction with a model-based observer.,1669
on drive vehicle. algorithm a tests developed wheel The experimental is rear through evaluated,The developed algorithm is evaluated through experimental tests on a rear wheel drive vehicle.,1670
role information. features efficiently temporal visual Extracting representation and a sequence plays in pivotal understanding,Extracting temporal and representation features efficiently plays a pivotal role in understanding visual sequence information.,1671
modeling out complexity the the rules explicitly it. of The of distribution possibility conditional,The complexity of the conditional distribution rules out the possibility of explicitly modeling it.,1672
"on optimization. exploit Instead, that the fact deep rely learning frameworks stochastic we","Instead, we exploit the fact that deep learning frameworks rely on stochastic optimization.",1673
study finite-sample Monte Carlo We means simulations. by of properties of the our estimators,We study the finite-sample properties of our estimators by means of Monte Carlo simulations.,1674
samples over consecutive accuracy identification suggest for promising The . days collected results,The results suggest promising identification accuracy for samples collected over consecutive days .,1675
"the ShelfNet. performance of experiments, validated we Through superior extensive","Through extensive experiments, we validated the superior performance of ShelfNet.",1676
This data use keypoint unlabeled train detector. the design extensively to the allows us to,This design allows us to extensively use the unlabeled data to train the keypoint detector.,1677
approach human this We structures. with empirically of block participants' ratings the assess,We empirically assess this approach with human participants' ratings of the block structures.,1678
node comprehensive enable to descriptors action over based joint on contextual segmentation inference information. as,as node descriptors to enable action segmentation based on joint inference over comprehensive contextual information.,1679
"do stored, With need or more. communicated, not calculated, any matched to descriptors this, be","With this, descriptors do not need to be calculated, stored, communicated, or matched any more.",1680
alternatives. to evaluate state-of-the-art relative performance We,We evaluate performance relative to state-of-the-art alternatives.,1681
their neural behavior. of (DNNs) networks explain is Understanding internal crucal the to representations deep,Understanding the internal representations of deep neural networks (DNNs) is crucal to explain their behavior.,1682
We show to reducible learning. is problem and the `corrective' `natural',We show the problem is reducible to `natural' and `corrective' learning.,1683
have super-resolution. networks Convolutional recently image demonstrated for interesting results single neural,Convolutional neural networks have recently demonstrated interesting results for single image super-resolution.,1684
"to with networks However, images. super-resolution deal were these on trained natural problem","However, these networks were trained to deal with super-resolution problem on natural images.",1685
"single present text our we for the super-resolution. database To network, evaluate image","To evaluate the network, we present our database for single text image super-resolution.",1686
selection model semiparametric studies This models. econometric paper in,This paper studies model selection in semiparametric econometric models.,1687
demonstrate LM-BIC simulations. and UT proposed procedures performance in The good semiparametric,The proposed semiparametric LM-BIC and UT procedures demonstrate good performance in simulations.,1688
"annotation. In Lucas-Kanade contrast, algorithm well the works without any (LK)","In contrast, the Lucas-Kanade (LK) algorithm works well without any annotation.",1689
"learned features. we perform on tracking LK testing, During the the","During testing, we perform the LK tracking on the learned features.",1690
"of such method contrast, In not annotation. any use does make our","In contrast, our method does not make use of any such annotation.",1691
"of calcium part insulin the is pancreatic secretion. glucose-stimulated In $\beta$-cells, signal oscillation core","In pancreatic $\beta$-cells, calcium oscillation signal is the core part of glucose-stimulated insulin secretion.",1692
plays part a in ATP process. this crucial,ATP plays a crucial part in this process.,1693
proposed providing capability provided and also Qualitative the analysis DiGGAN's has evidences. been demonstrates in,Qualitative analysis has also been provided and demonstrates the proposed DiGGAN's capability in providing evidences.,1694
Printed OCR an text is industrial systems. problem important recognition for,Printed text recognition is an important problem for industrial OCR systems.,1695
fashion settings. in is text standard in constructed Printed a procedural most,Printed text is constructed in a standard procedural fashion in most settings.,1696
understanding. is vital and Accurate tracking of detection effective objects for video,Accurate detection and tracking of objects is vital for effective video understanding.,1697
"be and impact does quality quality How marketplace predicted? outcomes, image automatically can","How does image quality impact marketplace outcomes, and can quality be automatically predicted?",1698
captioning networks image Deep the neural have great on successes task. achieved,Deep neural networks have achieved great successes on the image captioning task.,1699
generate plausible used how to to The teach captioning sentence corpus is the sentences. model,The sentence corpus is used to teach the captioning model how to generate plausible sentences.,1700
"paper, robust adversarially fast In training study of we this models.","In this paper, we study fast training of adversarially robust models.",1701
"solution. label, closed-form adversarial heuristic To the generate an we derive","To generate the adversarial label, we derive an closed-form heuristic solution.",1702
"coupled our model adversarial Then part, improves results. with state-of-the-art the the significantly label","Then coupled with the adversarial label part, our model significantly improves the state-of-the-art results.",1703
"dialogue, Argumentation-based been have handle which and exchange through arguments widely systems, can dialogue researched.","Argumentation-based dialogue systems, which can handle and exchange arguments through dialogue, have been widely researched.",1704
systems are dialogue missing dialogue exhaustively that information-seeking information. examine strategies based Existing handcrafted on,Existing information-seeking dialogue systems are based on handcrafted dialogue strategies that exhaustively examine missing information.,1705
"in arguments. However, strategies information not are rational specialized proposed constructing the for collecting","However, the proposed strategies are not specialized in collecting information for constructing rational arguments.",1706
two argumentative We of dialogue. domains conducted different dialogue experiments from using datasets,We conducted dialogue experiments using two datasets from different domains of argumentative dialogue.,1707
deep Recent impact learning successes neuroscience. have started in to,Recent successes in deep learning have started to impact neuroscience.,1708
perform demonstrate to poorly segmentation the We challenge. neuron first that state-of-the-art current approaches on,We first demonstrate that current state-of-the-art approaches to neuron segmentation perform poorly on the challenge.,1709
"run recognition It I-frames, need to for only i.e. complete a network","It only need to run a complete recognition network for I-frames, i.e.",1710
"knowledge planning procedure surface about the requires terrains. planetary no prior Moreover,","Moreover, the planning procedure requires no prior knowledge about planetary surface terrains.",1711
"dataset past, focused without we text image visual data. search in the only In","In the past, we only focused visual search in image dataset without text data.",1712
"search, of there Compared needs not has researches. multimodal been the sufficient to","Compared to the needs of multimodal search, there has not been sufficient researches.",1713
two image and tasks search as tasks. matching different text This regards two of,This regards two tasks of image search and text matching as two different tasks.,1714
compare we single these space searching Then datasets. multimodal to searching the performance in of,Then we compare the performance of multimodal searching to single space searching in these datasets.,1715
in localize objects advances to Recent attention of recognition maps utilize fine-grained interest.,Recent advances in fine-grained recognition utilize attention maps to localize objects of interest.,1716
attention by and activations. aggregating map max-pooling output The is generated the,The attention map is generated by aggregating and max-pooling the output activations.,1717
"model heavily i.e. state-of-the-art Our datasets, on results benchmarked achieves three","Our model achieves state-of-the-art results on three heavily benchmarked datasets, i.e.",1718
of information becoming machine is learning the various modes Incorporating a procedure into new trend.,Incorporating various modes of information into the machine learning procedure is becoming a new trend.,1719
algorithms way. detection proposed a saliency coarse-to-fine the multimodal on We validate in,We validate the proposed algorithms on multimodal saliency detection in a coarse-to-fine way.,1720
The weighting results adaptive of domain. maps each via from can be obtained final,The final results can be obtained via adaptive weighting of maps from each domain.,1721
They a part visual of significant human perception. embody,They embody a significant part of human visual perception.,1722
"gray-scale In given a hallucinating this paradigm of the image. from we explore colours paper,","In this paper, we explore the paradigm of hallucinating colours from a given gray-scale image.",1723
methods vividly through Our and of display colourization versatility proficiency life-like results outcomes. our the,Our results vividly display the versatility and proficiency of our methods through life-like colourization outcomes.,1724
"interplay between coexistence at sharing large unclear. a scale and inter-technology the remains Nevertheless, intra-network","Nevertheless, the interplay between intra-network sharing and inter-technology coexistence at a large scale remains unclear.",1725
"model to UNB and framework this analyze develop an analytical we networks. in paper, Thus,","Thus, in this paper, we develop an analytical framework to model and analyze UNB networks.",1726
optimal morbidity. stimulation the reduce of times surgical and Pre-operative can localization zone,Pre-operative localization of the optimal stimulation zone can reduce surgical times and morbidity.,1727
results with surgical our of planning. method a We compare registration-based,We compare our results with a registration-based method of surgical planning.,1728
directed studies process. demonstrate ML-guided that case the evolution provide then two We,We then provide two case studies that demonstrate the ML-guided directed evolution process.,1729
niche the tissue epithelial of and critical cell stem The to function. is morphology maintenance,The maintenance of the stem cell niche is critical to epithelial tissue morphology and function.,1730
this goal. We new this achieves demonstrate methodology that,We demonstrate that this new methodology achieves this goal.,1731
image same image large the use or typically patches. filters over the whole over They,They typically use the same filters over the whole image or over large image patches.,1732
local for they estimate then scale Only to perspective distortion. do compensate,Only then do they estimate local scale to compensate for perspective distortion.,1733
and function the can inherently utility between This trade-off demonstrate OMA NOMA.,This utility function can inherently demonstrate the trade-off between OMA and NOMA.,1734
"the Moreover, allocation derived. power in closed-form is the for solution second step","Moreover, the closed-form solution for power allocation in the second step is derived.",1735
accepted explanations Small understand. decision generally makers easier human are to as for,Small explanations are generally accepted as easier for human decision makers to understand.,1736
This paper constraint-agnostic any develops a computing solution for model. ML for explanations,This paper develops a constraint-agnostic solution for computing explanations for any ML model.,1737
of imagery can the achieved tweaking be interpolation effects control A coefficients. smooth by,A smooth control of imagery effects can be achieved by tweaking the interpolation coefficients.,1738
"even to the observed m-phases). be is types This same intersection within (n-arm,","This is even to be observed within the same intersection types (n-arm, m-phases).",1739
"By a requiring faster. deep is contrast, network, in single forward inference much a pass,","By contrast, inference in a deep network, requiring a single forward pass, is much faster.",1740
the There models. common two resulting optimizing of ways are,There are two common ways of optimizing the resulting models.,1741
is It gradient that methods known exhibit problems. well convergence ill-conditioned slow for based,It is well known that gradient based methods exhibit slow convergence for ill-conditioned problems.,1742
"networks with In similar we structure. scheme, use neural our two","In our scheme, we use two neural networks with similar structure.",1743
results show generate mask. our instance can Experimental fine-grained scheme that,Experimental results show that our scheme can generate fine-grained instance mask.,1744
also We a proposed method. lymphocyte segmentation,We also proposed a lymphocyte segmentation method.,1745
detection an nucleus also and We classification. segmentation lymphocyte proposed based on for algorithm,We also proposed an algorithm for lymphocyte segmentation based on nucleus detection and classification.,1746
paper. two major contributions are There this in,There are two major contributions in this paper.,1747
video tasks recognition activity most analysis. of in the challenging one is Egocentric,Egocentric activity recognition is one of the most challenging tasks in video analysis.,1748
requires a of and objects their discrimination manipulation. fine-grained It small,It requires a fine-grained discrimination of small objects and their manipulation.,1749
"images images, less potentially unlabelled than these useful Whilst of labelled can information. value contain","Whilst of less value than labelled images, these unlabelled images can contain potentially useful information.",1750
Current service a in genome (e.g. techniques a provider allow sequencing,Current techniques in sequencing a genome allow a service provider (e.g.,1751
segmentation a computer are in object for as existing methods Most vision formulated task. labeling,Most existing methods for object segmentation in computer vision are formulated as a labeling task.,1752
classical And the methods. compare result of with methods labeling these we some image,And we compare the result of these methods with some classical image labeling methods.,1753
"separately via framework. ""learning the translation"" undertakes components More two specifically, in SPGAN the","More specifically, SPGAN separately undertakes the two components in the ""learning via translation"" framework.",1754
networks. model re-ID using then existing a It learns,It then learns a re-ID model using existing networks.,1755
"re-ID integrates eSPGAN learning. seamlessly translation image and comparison, In model","In comparison, eSPGAN seamlessly integrates image translation and re-ID model learning.",1756
"cues. and Such photographs texture include which shading color, systems on been have trained","Such systems have been trained on photographs which include color, texture and shading cues.",1757
that results the show methods Simulation near-optimal performance. proposed yield,Simulation results show that the proposed methods yield near-optimal performance.,1758
A is in MagicVO shown the pipeline of Fig.,A pipeline of the MagicVO is shown in Fig.,1759
shuffle using operations). depth-wise convolutions or channel,using depth-wise convolutions or channel shuffle operations).,1760
exploit low-rank the matrices optimization a problem. regularized We combined to property formulate of the,We exploit the low-rank property of the matrices to formulate a combined regularized optimization problem.,1761
context single recovery. algorithm multi-channel We demonstrate images the MR of in our and,We demonstrate our algorithm in the context of single and multi-channel MR images recovery.,1762
that provides state-of-the-art approaches. proposed Numerical recovery algorithm experiments over improved indicate performance the the,Numerical experiments indicate that the proposed algorithm provides improved recovery performance over the state-of-the-art approaches.,1763
of language One survivors stroke have third difficulties.,One third of stroke survivors have language difficulties.,1764
with Our are in line lesion findings studies. previous,Our findings are in line with previous lesion studies.,1765
trained is additional without The Deep Regionlets framework efforts. end-to-end,The Deep Regionlets framework is trained end-to-end without additional efforts.,1766
effects. our view-dependent core is the component A of handling work of,A core component of our work is the handling of view-dependent effects.,1767
object. the input As of video we are an data RGB using,As input data we are using an RGB video of the object.,1768
of used geometry multi-view video stereo. reconstruct This to object a proxy via the is,This video is used to reconstruct a proxy geometry of the object via multi-view stereo.,1769
"propose we To a neural deep end, view-dependent network that EffectsNet, this effects. predicts","To this end, we propose EffectsNet, a deep neural network that predicts view-dependent effects.",1770
"images estimations, convert to to on these able images. diffuse are we observed Based","Based on these estimations, we are able to convert observed images to diffuse images.",1771
other into can images be views. These projected diffuse,These diffuse images can be projected into other views.,1772
"view, view-dependent the effects. the In new reinserts pipeline our target","In the target view, our pipeline reinserts the new view-dependent effects.",1773
based a multiple-kernel kernels on propose from gradients. We local-patch efficient match descriptor pixel,We propose a multiple-kernel local-patch descriptor based on efficient match kernels from pixel gradients.,1774
effect We patch whitening its and the the meaning. on semantic analyze demonstrate similarity of,We analyze the effect of the whitening on patch similarity and demonstrate its semantic meaning.,1775
the best is the without variant performing Our constructed of data. labeled descriptor unsupervised need,Our unsupervised variant is the best performing descriptor constructed without the need of labeled data.,1776
"factors the with relate is observation. $y_{n}=Qf_{n}+\varepsilon_{n}$ the The taken to equation,","The equation, $y_{n}=Qf_{n}+\varepsilon_{n}$ is taken to relate the factors with the observation.",1777
rank of Estimating determining $p$ numerical to $\tilde{M}$. is the equated,Estimating $p$ is equated to determining the numerical rank of $\tilde{M}$.,1778
"groups In addition, learn among the improve representational to we effective capability. connections","In addition, we learn effective connections among groups to improve the representational capability.",1779
"other tasks. Group-Net generalization proposed the to shows Moreover, strong","Moreover, the proposed Group-Net shows strong generalization to other tasks.",1780
process. training network such challenge non-differentiable The in top-K is the a selection,The challenge in training such a network is the non-differentiable top-K selection process.,1781
to an obtain visually Braille information. impaired and learn is for way the effective knowledge,Braille is an effective way for the visually impaired to learn knowledge and obtain information.,1782
to image Braille detect dots the image. in Braille whole aims Braille recognition automatically,Braille image recognition aims to automatically detect Braille dots in the whole Braille image.,1783
and dots the is basic Braille image core for detection step Braille recognition.,Braille dots detection is the core and basic step for Braille image recognition.,1784
have website. released We images our GitHub dataset on Braille the,We have released our Braille images dataset on the GitHub website.,1785
use tree a introduce novel policy guide to algorithms that search search. two We,We introduce two novel tree search algorithms that use a policy to guide search.,1786
problems. this for particularly best-first `needle-in-a-haystack' that suited show We well is algorithm,We show that this best-first algorithm is particularly well suited for `needle-in-a-haystack' problems.,1787
range for the recovery that We ranks. show of holds a wide multilinear correct,We show that the correct recovery holds for a wide range of multilinear ranks.,1788
"When and set-point its HIV viral epidemics, load is heritability. incorporate important modelling to it","When modelling HIV epidemics, it is important to incorporate set-point viral load and its heritability.",1789
trained Recently been it has clean also methods without such that targets. can shown be,Recently it has been shown that such methods can also be trained without clean targets.,1790
"does clean images. It nor target noisy require image pairs, not","It does not require noisy image pairs, nor clean target images.",1791
"captured a lack result, such As under images conditions contrast.","As a result, images captured under such conditions lack contrast.",1792
performs existing well that Evaluations our state-of-the-art show methods. compared model the to,Evaluations show that our model performs well compared to the existing state-of-the-art methods.,1793
model It various diverse in of situations verifies conditions. also and our the lighting potential,It also verifies the potential of our model in diverse situations and various lighting conditions.,1794
connections Modern certain networks alleviate implement shortcut generally form deep difficulties. optimization to of a,Modern deep networks generally implement a certain form of shortcut connections to alleviate optimization difficulties.,1795
"network topology that of such we alters the deep However, nature observe networks.","However, we observe that such network topology alters the nature of deep networks.",1796
"to many aggregated In wide these networks. similarly networks ways, behave","In many ways, these networks behave similarly to aggregated wide networks.",1797
works time. practice This in most well,This practice works well in most time.,1798
phenomena the problem. this as sparse We early gradients term,We term this phenomena as the early sparse gradients problem.,1799
"magnitude this In and can the gradients. of avoid increase explosion we way, distance","In this way, we can avoid distance explosion and increase the magnitude of gradients.",1800
Omniglot conducted Extensive datasets. experiments and are on miniImageNet,Extensive experiments are conducted on Omniglot and miniImageNet datasets.,1801
recurrent type layer most Term is of used Long-Short Memory The the (LSTM).,The most used type of recurrent layer is the Long-Short Term Memory (LSTM).,1802
and tone adjustments. color themes with unique Color enhance and stylization tone to artistic strives,Color and tone stylization strives to enhance unique themes with artistic color and tone adjustments.,1803
"hand, On adjustments various to stylistic needs enhancement the regions. semantic other distinct to apply","On the other hand, stylistic enhancement needs to apply distinct adjustments to various semantic regions.",1804
broader a enables an ability of range styles. visual Such,Such an ability enables a broader range of visual styles.,1805
our Image accomplished can efficiently deep be a with through pass network. single stylization forward,Image stylization can be efficiently accomplished with a single forward pass through our deep network.,1806
effectiveness datasets our demonstrate learning architecture. for image of the deep existing Experiments stylization on,Experiments on existing datasets for image stylization demonstrate the effectiveness of our deep learning architecture.,1807
"along spatial dimension well. Similarly, GSoP scaling perform for we tensor can as","Similarly, we can perform GSoP along spatial dimension for tensor scaling as well.",1808
to we this on use adapt In estimation work architecture pose edge multi-person it devices.,In this work we adapt multi-person pose estimation architecture to use it on edge devices.,1809
and a are model available part of The OpenVINO$^{TM}$ as Intel$\unicode{xAE}$ code Toolkit.,The code and model are available as a part of Intel$\unicode{xAE}$ OpenVINO$^{TM}$ Toolkit.,1810
work This first segmentation of with data. a this introduces for semantic kind baseline,This work introduces a first baseline for semantic segmentation with this kind of data.,1811
this discussed. of are possible causes phenomenon The,The possible causes of this phenomenon are discussed.,1812
"are regions Quite an often, interesting most these image. of the parts","Quite often, these regions are the most interesting parts of an image.",1813
captured this Almost every today is of way. interest event,Almost every event of interest today is captured this way.,1814
the propose to a comprehensive problem. solution We,We propose a comprehensive solution to the problem.,1815
"based combines appearance are solution and proximity. cues geometry, that Our on","Our solution combines cues that are based on geometry, appearance and proximity.",1816
"and to set new, challenging, collected algorithm. evaluate We a data our","We collected a new, and challenging, data set to evaluate our algorithm.",1817
al. Beutner et This in results contrasts,This contrasts results in Beutner et al.,1818
consistently layer it. Our without the improves the spatial performance attention model over,Our spatial attention layer consistently improves the performance over the model without it.,1819
methods real in these applied effect is on of world unknown. images The the static,The effect of these methods applied on static images in the real world is unknown.,1820
"comparative have also effectiveness extensive experiments demonstrated components. of fully Besides, different the","Besides, extensive comparative experiments have also fully demonstrated the effectiveness of different components.",1821
"problem two break system with devices, down. different Consider a which can reliability the in","Consider the reliability problem in a system with two different devices, which can break down.",1822
estimation. on keypoint We human a based system sign language translation propose,We propose a sign language translation system based on human keypoint estimation.,1823
"study recognition how networks to subject important are it face to is attacks. So,","So, it is important to study how face recognition networks are subject to attacks.",1824
samples is spectral estimation utilizing from (LSE) Line methods. studied variational multi snapshot Bayesian the,Line spectral estimation (LSE) from multi snapshot samples is studied utilizing the variational Bayesian methods.,1825
"of state-of-the-art competitive numerical demonstrate to compared methods. results the the performance Finally, MVALSE","Finally, numerical results demonstrate the competitive performance of the MVALSE compared to state-of-the-art methods.",1826
"salient, becomes As less difficult. more a objects identifying relatively target itself referred becomes","As a target becomes relatively less salient, identifying referred objects itself becomes more difficult.",1827
our effectiveness Experimental show results the of approach.,Experimental results show the effectiveness of our approach.,1828
writing typical skills. and Developmental the to inability acquire reading dyslexia is characterized by,Developmental dyslexia is characterized by the inability to acquire typical reading and writing skills.,1829
were that derived hypotheses previous findings. tested these We from two directly here,We here tested two hypotheses that were directly derived from these previous findings.,1830
"RANln scores. with hypothesized we left Second, the dyslexics connectivity mPT-MGB correlates amount that of","Second, we hypothesized that the amount of left mPT-MGB connectivity correlates with dyslexics RANln scores.",1831
"the without to data, and of parameter tuned. number be any","the number of data, and without any parameter to be tuned.",1832
only detection presents paper supervision. change novel This weak semantic with scene scheme a,This paper presents a novel semantic scene change detection scheme with only weak supervision.,1833
"this specific for becomes and usually dataset labor-intensive indispensable. task, a which is time-consuming, However,","However, a specific dataset for this task, which is usually labor-intensive and time-consuming, becomes indispensable.",1834
collections is trained Our truth. ground using image network without large uncontrolled,Our network is trained using large uncontrolled image collections without ground truth.,1835
"By can incorporating our renderer, network a self-supervision. differentiable from learn","By incorporating a differentiable renderer, our network can learn from self-supervision.",1836
supervision depth provides map normal estimation. MVS direct for coarse also,MVS depth also provides direct coarse supervision for normal map estimation.,1837
attempt first use rendering. is to We MVS supervision for learning believe this inverse the,We believe this is the first attempt to use MVS supervision for learning inverse rendering.,1838
"is images, floor like the plans. area in there document no such However, attempt of","However, there is no such attempt in the area of document images, like floor plans.",1839
"earlier. from plan sketches, synthesis were proposed well Floor data-driven as as models,","Floor plan synthesis from sketches, as well as data-driven models, were proposed earlier.",1840
description building images floor to first plan from attempt automatically. textual the is render Ours,Ours is the first attempt to render building floor plan images from textual description automatically.,1841
publicly We available on benchmark plan floor experimented have datasets.,We have experimented on publicly available benchmark floor plan datasets.,1842
are capacity for Current range. systems currently optimized radio and,Current radio systems are currently optimized for capacity and range.,1843
"reliable applications certain of require fast communication wireless systems over and However, distances. short","However, certain applications of wireless systems require fast and reliable communication over short distances.",1844
There performance using limitations are of RELU. just,There are performance limitations of just using RELU.,1845
convolutional all RELU for used is layers.,RELU is used for all convolutional layers.,1846
each of views. architecture of two two network Our branches encoder-decoder the analyzing for consists,Our network architecture consists of two encoder-decoder branches for analyzing each of the two views.,1847
significantly approaches based a dataset. projection outperform synthetic large-scale We existing the on,We significantly outperform the existing projection based approaches on a large-scale synthetic dataset.,1848
"To we novel multi-path adaptive propose (MAMNet). modulation a address issue, this network","To address this issue, we propose a novel multi-path adaptive modulation network (MAMNet).",1849
communicating to mainly effective harder time-dependent of This devices. owing the qualities link is the,This is harder mainly owing to the time-dependent effective link qualities of the communicating devices.,1850
wireless of the components improve systems and . to,of wireless systems and to improve the components .,1851
probabilistic introduced feature background learns distributions. The a module and target appearance generative model of,The introduced appearance module learns a probabilistic generative model of target and background feature distributions.,1852
the segmentation Comprehensive effectiveness approach demonstrate experiments object video three proposed of on the benchmarks.,Comprehensive experiments demonstrate the effectiveness of the proposed approach on three video object segmentation benchmarks.,1853
"method large-scale YouTube-VOS Furthermore, on our dataset. the approaches outperforms published all","Furthermore, our method outperforms all published approaches on the large-scale YouTube-VOS dataset.",1854
"by biological which recover tissues paper, the this investigate mechanisms we after injury. In","In this paper, we investigate the mechanisms by which biological tissues recover after injury.",1855
mobile for device chip enables and models within one IoT applications. It multiple single,It enables multiple models within one single chip for mobile and IoT device applications.,1856
transferability We action recognition to systems. black-box of demonstrate the also attacks our,We also demonstrate the transferability of our attacks to black-box action recognition systems.,1857
efficient the of (PMBM) multi-Bernoulli an trajectory implementation filter. paper mixture Poisson proposes This,This paper proposes an efficient implementation of the Poisson multi-Bernoulli mixture (PMBM) trajectory filter.,1858
"require computerized procedures extensive However, the planning.","However, the procedures require extensive computerized planning.",1859
normal training network unpaired images real The using is samples maps. trained rendered of and,The network is trained using unpaired training samples of real images and rendered normal maps.,1860
sketch drawing in Recognizing free-hand applications. real-world extremely useful many the is,Recognizing the free-hand sketch drawing is extremely useful in many real-world applications.,1861
"in instance-level previous limitations key However, several works. are SBIR there","However, there are several key limitations in previous instance-level SBIR works.",1862
"limitations break in our proposed previous existed can DeepTCNet Critically, works. the","Critically, our proposed DeepTCNet can break the limitations existed in previous works.",1863
proposed five benchmark model. the sketch on of The extensive the experiments datasets effectiveness validate,The extensive experiments on five benchmark sketch datasets validate the effectiveness of the proposed model.,1864
show benchmark competitive our perform models datasets. that results on We can,We show that our models can perform competitive results on benchmark datasets.,1865
source We models available. publicly of make our code the,We make the source code of our models publicly available.,1866
pose-guided new a human of to deep approach We resynthesis photographs. present learning,We present a new deep learning approach to pose-guided resynthesis of human photographs.,1867
with is the estimated field fully-convolutional skip deformable connections guided used. correspondence A architecture by,A fully-convolutional architecture with deformable skip connections guided by the estimated correspondence field is used.,1868
We show state-of-the-art result for image pose-guided synthesis.,We show state-of-the-art result for pose-guided image synthesis.,1869
"the transfer pose-guided system and face for resynthesis. we performance our of garment demonstrate Additionally,","Additionally, we demonstrate the performance of our system for garment transfer and pose-guided face resynthesis.",1870
estimator minimax under is the a It optimal Nikodym metric.,It is a minimax optimal estimator under the Nikodym metric.,1871
provide We efficient $\widehat{\mathcal{C}}$. algorithm on compute then an projection to,We then provide an efficient algorithm to compute projection on $\widehat{\mathcal{C}}$.,1872
good empirical simulations. We a behavior method observe the of in,We observe a good empirical behavior of the method in simulations.,1873
"Further, provide the by Chebyshev we the analysis variation of characterization approximation.","Further, we provide analysis of the variation characterization by the Chebyshev approximation.",1874
performance. then We two recognition complementary for improving propose early approaches,We then propose two complementary approaches for improving early recognition performance.,1875
"images medical {\em phases real-world complementary acquired In different possess e.g. information, practice, in","In real-world practice, medical images acquired in different phases possess complementary information, {\em e.g.",1876
two-phase This studies in paper scans. CT segmentation organ,This paper studies organ segmentation in two-phase CT scans.,1877
can PCN jointly. relations phase-to-phase and formulate mathematically be to explained data-to-label,PCN can be mathematically explained to formulate phase-to-phase and data-to-label relations jointly.,1878
"public PCN transfers to datasets, applications. potential single-phase its two well demonstrating Besides,","Besides, PCN transfers well to two public single-phase datasets, demonstrating its potential applications.",1879
"a We i.e. burst as pipeline on photography focus the both whole,","We focus both on the burst photography pipeline as a whole, i.e.",1880
"as as and task. well Gaussian demosaicking traditional denoising, on denoising burst the","burst demosaicking and denoising, as well as on the traditional Gaussian denoising task.",1881
"vector programs methods in this paper, we In spaces. of introduce propositional encoding logic","In this paper, we introduce methods of encoding propositional logic programs in vector spaces.",1882
are vectors by by Interpretations and matrices. programs are represented represented,Interpretations are represented by vectors and programs are represented by matrices.,1883
"challenging address this task, propose new Siamese R-CNN. To Mask we","To address this challenging new task, we propose Siamese Mask R-CNN.",1884
Our an proposed end-to-end be can trained MGG fashion. in,Our proposed MGG can be trained in an end-to-end fashion.,1885
remarkable both attention industry has driving Autonomous attracted and from academia.,Autonomous driving has attracted remarkable attention from both industry and academia.,1886
fitting improves We using keypoints that significantly performance. show,We show that using keypoints significantly improves fitting performance.,1887
further suggest we By human performance comparing future improvements. with for several directions,By comparing with human performance we suggest several future directions for further improvements.,1888
"simple, intuitive, very contains parameters. and is Our algorithm very few","Our algorithm is very simple, intuitive, and contains very few parameters.",1889
of evaluate car dataset. approach our the VehicleSemanticPart in the subclass We,We evaluate our approach in the car subclass of the VehicleSemanticPart dataset.,1890
of multiple-choice with VQA and proposed VQA models deal are Two to respectively. types open-ended,Two types of models are proposed to deal with open-ended VQA and multiple-choice VQA respectively.,1891
three benchmarks. VQA We evaluate models on our,We evaluate our models on three VQA benchmarks.,1892
method. demonstrates the effectiveness state-of-the-art proposed the comparable with performance the of The,The comparable performance with the state-of-the-art demonstrates the effectiveness of the proposed method.,1893
one in hand language. sign The methods used of typical the are gestures,The hand gestures are one of the typical methods used in sign language.,1894
with world. communicate the It to the for people very hearing-impaired is difficult,It is very difficult for the hearing-impaired people to communicate with the world.,1895
"will soundtrack the recorded the on Based be gestures, recognized played.","Based on the recognized gestures, the recorded soundtrack will be played.",1896
collection GECV is crowds containing dataset people. of videos of a,GECV dataset is a collection of videos containing crowds of people.,1897
"means, nonlocal plain Filtering patch-wise data. exclusively SAR by operating on performed is","Filtering is performed by plain patch-wise nonlocal means, operating exclusively on SAR data.",1898
reconstructions complexes Electron high-resolution nearnative allows state. tomography (ET) at of macromolecular,Electron tomography (ET) allows high-resolution reconstructions of macromolecular complexes at nearnative state.,1899
state-of-the-art grid is approach frameworks. The to different guided easy detection be to localization extended,The grid guided localization approach is easy to be extended to different state-of-the-art detection frameworks.,1900
"and Furthermore, patients. these changes of network impairment significantly the cognitive predicted memory","Furthermore, these network changes significantly predicted the cognitive and memory impairment of patients.",1901
structures evaluation Background Combined objective: lumbosacral (e.g. and of,Background and objective: Combined evaluation of lumbosacral structures (e.g.,1902
bone) on CT non-contrast & with learning. deep nerves,nerves & bone) on non-contrast CT with deep learning.,1903
of nerve results testing successful the lumbosacral and on reveals segmentation bone CT. Results:,Results: the testing results reveals successful segmentation of lumbosacral bone and nerve on CT.,1904
"way, related In naturally passed information ones. objects smaller larger to can this from","In this way, information can naturally passed from larger objects to smaller related ones.",1905
"been IoT codes Recently, non-orthogonal massive have for advocated access.","Recently, non-orthogonal codes have been advocated for IoT massive access.",1906
Our this addressing challenge towards contributions as follows. are,Our contributions towards addressing this challenge are as follows.,1907
"that recurrent the noise Here, networks. information can demonstrate improve in flux also neural we","Here, we demonstrate that noise can also improve the information flux in recurrent neural networks.",1908
"segmentation object and is in visual problems, detection. Contextual vital such semantic understanding information as","Contextual information is vital in visual understanding problems, such as semantic segmentation and object detection.",1909
"operation, dependencies. the further pixel capture recurrent By finally taking can a each full-image","By taking a further recurrent operation, each pixel can finally capture the full-image dependencies.",1910
plays from in a anti-spoofing role systems attacks. protecting recognition various face crucial Face,Face anti-spoofing plays a crucial role in protecting face recognition systems from various attacks.,1911
pedestrian from by to views re-identification Person detectors. detected pedestrian no-overlap is camera retrieval images,Person re-identification is to retrieval pedestrian images from no-overlap camera views detected by pedestrian detectors.,1912
in issue real usability significantly scalability and of limits models world. the the the This,This issue significantly limits the scalability and usability of the models in the real world.,1913
"task. re-identification we IBN-reID Furthermore, model for person the propose","Furthermore, we propose IBN-reID model for the person re-identification task.",1914
supervised IBN-reID model methods. translated The on by trained is the images,The IBN-reID model is trained on the translated images by supervised methods.,1915
achieved our by Very re-ID accuracy method. competitive is,Very competitive re-ID accuracy is achieved by our method.,1916
"methods a not of the range. most present do times good Also, capture these","Also, these methods most of the times do not present a good capture range.",1917
"of Neural propose deal with Convolutional this we the problem, To use Network.","To deal with this problem, we propose the use of Convolutional Neural Network.",1918
registration promising be The methods. compared traditional results to to image showed,The results showed to be promising compared to traditional image registration methods.,1919
representations target the in the An in source have domain. domain image might multiple,An image in the source domain might have multiple representations in the target domain.,1920
map space. both a on rely approaches mostly into simplifying shared-latent Current to assumptions domains,Current approaches mostly rely on simplifying assumptions to map both domains into a shared-latent space.,1921
"able the domain-invariant information model to between they two modalities. only Consequently, are the","Consequently, they are only able to model the domain-invariant information between the two modalities.",1922
individually. existing CNNs on samples the powerful methods transform Most rely to,Most existing methods rely on powerful CNNs to transform the samples individually.,1923
"in relations propose to consider transformation. we sample In contrast, the the","In contrast, we propose to consider the sample relations in the transformation.",1924
"achieve goal, spectral this we into To clustering CNN. technique incorporate","To achieve this goal, we incorporate spectral clustering technique into CNN.",1925
Bayesian by a also This be employing can derived result formulation.,This result can also be derived by employing a Bayesian formulation.,1926
"to that for paired edits attempted semantic required i.e. data, use Previous warping approaches","Previous approaches that attempted to use warping for semantic edits required paired data, i.e.",1927
of example subject semantic same different with images the attributes.,example images of the same subject with different semantic attributes.,1928
also preserving at better the subject's that show our edits identity. are substantially We,We also show that our edits are substantially better at preserving the subject's identity.,1929
to (DNNs) networks adversarial Deep be demonstrated been have to neural vulnerable examples.,Deep neural networks (DNNs) have been demonstrated to be vulnerable to adversarial examples.,1930
"the images to deep trained fool clean can well networks. imperceptible Specifically, perturbations adding neural","Specifically, adding imperceptible perturbations to clean images can fool the well trained deep neural networks.",1931
"model paper, \textbf{ComDefend}. end-to-end to adversarial propose examples: compression image we an In this defend","In this paper, we propose an end-to-end image compression model to defend adversarial examples: \textbf{ComDefend}.",1932
the reconstruct used with image is the quality. to original And ResCNN high,And the ResCNN is used to reconstruct the original image with high quality.,1933
"a named In this novel scene paper, we detection propose text TextMountain. method","In this paper, we propose a novel scene text detection method named TextMountain.",1934
border-center of TextMountain use full of The making idea information. key is,The key idea of TextMountain is making full use of border-center information.,1935
We for Indian Images. proposed the network Landmark ensemble classification an of purpose of,We proposed an ensemble network for the purpose of classification of Indian Landmark Images.,1936
adaptive networks. final learning architecture all the The is of mentioned an,The final architecture is an adaptive learning of all the mentioned networks.,1937
category eliminate produces false a score The reliable to cases. proposed network,The proposed network produces a reliable score to eliminate false category cases.,1938
images. computer method are vision nets leading neural (CNN) classifying Convolutional the for,Convolutional neural nets (CNN) are the leading computer vision method for classifying images.,1939
map the with image is combined net. This then classification,This map is then combined with the image classification net.,1940
approach the dataset was and OpenSurfaces evaluated materials COCO dataset. using the This object,This approach was evaluated using the COCO object dataset and the OpenSurfaces materials dataset.,1941
"results completely gave methods to In that background cases, both ignore the it superior region.","In both cases, it gave superior results to methods that completely ignore the background region.",1942
when for is for driving segmentation This example. autonomous semantic information using critical,This information is critical when using semantic segmentation for autonomous driving for example.,1943
have Standard semantic well-established segmentation metrics. evaluation systems,Standard semantic segmentation systems have well-established evaluation metrics.,1944
"model swapping. paper generative face image-based for This presents deep FSNet, a","This paper presents FSNet, a deep generative model for image-based face swapping.",1945
a signs time system images these is a detect Developing real to from challenge. great,Developing a real time system to detect these signs from images is a great challenge.,1946
"for Lung we segmentation Radiographs. Chest In paper, model a propose this in robust","In this paper, we propose a robust model for Lung segmentation in Chest Radiographs.",1947
"reason. have there studies no determine the systematic underlying to However, been","However, there have been no systematic studies to determine the underlying reason.",1948
a from breach. essential Face recognition to anti-spoofing security prevent systems face is,Face anti-spoofing is essential to prevent face recognition systems from a security breach.,1949
"often precision However, methods weights represent existing with and same all (bit-width). activations the quantization","However, existing quantization methods often represent all weights and activations with the same precision (bit-width).",1950
rotates and an The scans scene automated pan-tilt fashion. camera entire in RGB-D the,The pan-tilt RGB-D camera rotates and scans the entire scene in an automated fashion.,1951
"In between and local two scenes. registration, matched features extracted are","In local registration, features are extracted and matched between two scenes.",1952
"exceed distances to a correspondences, are rejected. trajectories threshold, rotation the False-positive whose","False-positive correspondences, whose distances to the rotation trajectories exceed a threshold, are rejected.",1953
"and Oftentimes, different appear to textures. due shape occlusion from similar symmetries, repetitive viewpoints objects","Oftentimes, objects appear similar from different viewpoints due to shape symmetries, occlusion and repetitive textures.",1954
explicitly we work uncertainty. with propose this to deal In this,In this work we propose to explicitly deal with this uncertainty.,1955
significant non-differentiable is algorithms. some A of impediment the nature,A significant impediment is the non-differentiable nature of some algorithms.,1956
We to propose use overcome difficulty. gradients (SG) to this synthetic,We propose to use synthetic gradients (SG) to overcome this difficulty.,1957
universal networks. the approximation property neural uses function SG of,SG uses the universal function approximation property of neural networks.,1958
more makes based processing a video This model still complex task.,This makes model based video processing a still more complex task.,1959
in an web essential instructional aspects. understanding of video two videos is branch Understanding,Understanding web instructional videos is an essential branch of video understanding in two aspects.,1960
"reasoning paper, (QA). study question-answering In this via on instructional we videos","In this paper, we study reasoning on instructional videos via question-answering (QA).",1961
"an video applications. community emphasis it in Surprisingly, has the its despite rich been not","Surprisingly, it has not been an emphasis in the video community despite its rich applications.",1962
image metric are Two content-based major learning retrieval applications and face of verification.,Two major applications of metric learning are content-based image retrieval and face verification.,1963
"training. non-parametric the tasks, majority the of retrieval triplet-based current state-of-the-art (SOTA) approaches are For","For the retrieval tasks, the majority of current state-of-the-art (SOTA) approaches are triplet-based non-parametric training.",1964
"For face approaches however, verification have recent SOTA parametric the adopted classification-based tasks, training.","For the face verification tasks, however, recent SOTA approaches have adopted classification-based parametric training.",1965
first the We as the a to tackle problem image work continuous introduce retrieval operation.,We introduce the first work to tackle the image retrieval problem as a continuous operation.,1966
investigated. has point relation However been rarely the views cloud and between,However the relation between point cloud and views has been rarely investigated.,1967
our state-of-the-art compared can results performance Experimental indicate model achieve improvement models. the significant with,Experimental results indicate our model can achieve significant performance improvement compared with the state-of-the-art models.,1968
of lossless for can The method network a used well. compression be as proposed neural,The proposed method can be used for lossless compression of a neural network as well.,1969
"the results However, care. their requires of interpretation","However, the interpretation of their results requires care.",1970
representational predictions comprehensive spaces. models make about Encoding,Encoding models make comprehensive predictions about representational spaces.,1971
include linear-model decoding Encoding fitted models and components. typically,Encoding and decoding models typically include fitted linear-model components.,1972
generalization in and of performance. terms their Encoding evaluated decoding are models,Encoding and decoding models are evaluated in terms of their generalization performance.,1973
achieves the of depends level (e.g. a The on interpretation model generalization correct,The correct interpretation depends on the level of generalization a model achieves (e.g.,1974
theoretical Many be for tested must drive progress. compared models analyses inferentially to and,Many models must be tested and inferentially compared for analyses to drive theoretical progress.,1975
videos and drastically (egocentric) are nature. in (exocentric) different third person First-person,First-person (egocentric) and third person (exocentric) videos are drastically different in nature.,1976
also relating two in aspects. two (egocentric explore the domains We and exocentric),We also explore relating the two domains (egocentric and exocentric) in two aspects.,1977
"the problem the cross-view Second, retrieval address views. two we across","Second, we address the cross-view retrieval problem across the two views.",1978
show retrieving be data. data real using synthetic beneficial could in that We,We show that using synthetic data could be beneficial in retrieving real data.,1979
are dataset code publicly The available. and,The code and dataset are publicly available.,1980
"However, practical resource-constrained in restricted environments. their often applications are","However, their practical applications are often restricted in resource-constrained environments.",1981
"for appropriate an Moreover, current provide such benchmarks evaluation cases. cannot","Moreover, current benchmarks cannot provide an appropriate evaluation for such cases.",1982
Source will dataset publicly available. and code be made,Source code and dataset will be made publicly available.,1983
vision including to many tasks Adversarial demonstrated threaten object been detection. computer examples have,Adversarial examples have been demonstrated to threaten many computer vision tasks including object detection.,1984
"this processing reduced. is way, the In time","In this way, the processing time is reduced.",1985
"representation flux"" major This approaches. two previous context advantages ""image over has","This ""image context flux"" representation has two major advantages over previous approaches.",1986
intelligence brought of forefront. rise artificial (AI) learning has The to deep the,The rise of deep learning has brought artificial intelligence (AI) to the forefront.,1987
weak strong all than to AI AI. achievements belong These rather,These achievements all belong to weak AI rather than strong AI.,1988
achieve to How science. AI yet strong intelligence known not of field the is in,How to achieve strong AI is not known yet in the field of intelligence science.,1989
"Cognitive field calling (TCR). a paradigm, Theory especially Relativity this new for is Currently, of","Currently, this field is calling for a new paradigm, especially Theory of Cognitive Relativity (TCR).",1990
derived two from principles facts life and are The experience. scientific,The two principles are derived from scientific facts and life experience.,1991
"different animals'. artificial consciousness, from intelligence general particularly or and humans' artificial","artificial general intelligence or artificial consciousness, particularly different from humans' and animals'.",1992
"and them. largely encoder, across decoder information improves can Which flow the","Which can largely improves the information flow encoder, decoder and across them.",1993
for accuracy. better in overfitting MDU-Net reduce Which,Which reduce overfitting in MDU-Net for better accuracy.,1994
rearrangements. are with diseases connected chromatin Many hereditary,Many hereditary diseases are connected with chromatin rearrangements.,1995
"those paper, solid we In develop method to Gregory models. based parameterize this a","In this paper, we develop a Gregory solid based method to parameterize those models.",1996
"trivariate First, the we to Gregory representation solid Gregory representation. patch extend the","First, we extend the Gregory patch representation to the trivariate Gregory solid representation.",1997
"not properly utilize potential of do the their recursive the Nevertheless, designs operation.","Nevertheless, their designs do not properly utilize the potential of the recursive operation.",1998
"methods. the outperforms that show we other state-of-the-art method our In addition,","In addition, we show that our method outperforms the other state-of-the-art methods.",1999
computer tasks. is Semantic problem many key segmentation vision a for,Semantic segmentation is a key problem for many computer vision tasks.,2000
We of the in verify two approach aspects. effectiveness our,We verify the effectiveness of our approach in two aspects.,2001
"we positive for with this generating guaranteed Jacobian. method paper, TBSs propose In a","In this paper, we propose a method for generating TBSs with guaranteed positive Jacobian.",2002
"is Finally, improved. TBSs smoothness between the adjacent","Finally, the smoothness between adjacent TBSs is improved.",2003
misalignment common the leads between accuracy. to the and This object localization final classification confidence,This leads to the common misalignment between the final object classification confidence and localization accuracy.,2004
"we problems. In to a paper, Transformer propose these RoI this address","In this paper, we propose a RoI Transformer to address these problems.",2005
Transformer. have and also experiments of our flexibility the effectiveness validated Extensive RoI,Extensive experiments have also validated the flexibility and effectiveness of our RoI Transformer.,2006
simultaneously continuously. and models strength arguments' update Continuous,Continuous models update arguments' strength simultaneously and continuously.,2007
discrete several models solutions. contains and to compute implementations of algorithms continuous numerical and Attractor,Attractor contains implementations of several discrete and continuous models and numerical algorithms to compute solutions.,2008
"covering of a dataset variety by three simultaneously is recorded wide scenarios. cameras, Our real-life","Our dataset is recorded simultaneously by three cameras, covering a wide variety of real-life scenarios.",2009
"involve methods parameter However, current plausible considerable to generate tweaking behaviors.","However, current methods involve considerable parameter tweaking to generate plausible behaviors.",2010
"algorithm optimization and constraints, collision velocity attraction, direction considers several including control. continuity, Our avoidance,","Our optimization algorithm considers several constraints, including velocity continuity, collision avoidance, attraction, and direction control.",2011
signal a from speed way detector. slow CTGI indicates detect to a high non-reproducible,CTGI indicates a way to detect high speed non-reproducible signal from a slow detector.,2012
"is conducted to achieve scene. areas of the parallelly CTGI this, To different","To achieve this, CTGI is conducted parallelly to different areas of the scene.",2013
imaging. can have This method in strong ultrafast application a,This method can have a strong application in ultrafast imaging.,2014
(VQA) content. questions `Known' answer systems intelligent Current about Question Answering Visual can visual,Current Visual Question Answering (VQA) systems can answer intelligent questions about `Known' visual content.,2015
"for untrimmed streams. long, natural retrieval strives language video research in moment This","This research strives for natural language moment retrieval in long, untrimmed video streams.",2016
misalignment. semantic and crucial two misalignment structural We identify challenges:,We identify two crucial challenges: semantic misalignment and structural misalignment.,2017
"accurate of is and great classification Therefore, arterioles an venules of necessity.","Therefore, an accurate classification of arterioles and venules is of great necessity.",2018
color the network inputs outputs. fundus takes as labels multi-class image original This as and,This network takes the original color fundus image as inputs and multi-class labels as outputs.,2019
our structure of (Unet) backbone the adopt network the as model. encoding-decoding We proposed,We adopt the encoding-decoding structure (Unet) as the backbone network of our proposed model.,2020
"and The filter recovery, maximum banks achieve orthogonality. decimation, perfect","The filter banks achieve maximum decimation, perfect recovery, and orthogonality.",2021
uses domain. transform the frequency The graph proposed in sampling,The proposed transform uses sampling in the graph frequency domain.,2022
reconstruction conditions examples. and design the perfect clarify show We,We clarify the perfect reconstruction conditions and show design examples.,2023
"no input data InGAN image fully the unsupervised, other than itself. is requiring additional","InGAN is fully unsupervised, requiring no additional data other than the input image itself.",2024
"down-stream tasks. original to space for the are After coordinate features relation-aware distributed back reasoning,","After reasoning, relation-aware features are distributed back to the original coordinate space for down-stream tasks.",2025
"abundant. the data, much is hand, on more Unlabeled other","Unlabeled data, on the other hand, is much more abundant.",2026
"such a as unlabeled could to data. co-training, learning powerful leverage provide tool techniques, Semi-supervised","Semi-supervised learning techniques, such as co-training, could provide a powerful tool to leverage unlabeled data.",2027
"can advantage our of take training. during data approach unlabeled Thus,","Thus, our approach can take advantage of unlabeled data during training.",2028
"are video Extensive large-scale two benchmarks, conducted experiments on and FCVID ActivityNet.","Extensive experiments are conducted on two large-scale video benchmarks, FCVID and ActivityNet.",2029
hardly compatible problem search architectures for is are The the CNN self-similarities. that with,The problem is that CNN architectures are hardly compatible with the search for self-similarities.,2030
CNN clean a presented which trained the to predict is This is to information image.,This information is presented to a CNN which is trained to predict the clean image.,2031
image architecture apply proposed We and denoising. to the video,We apply the proposed architecture to image and video denoising.,2032
recognition gesture radar. simple We introduce technique hand effective in a but automatic using,We introduce a simple but effective technique in automatic hand gesture recognition using radar.,2033
classifies on micro-Doppler gestures envelopes of the their based signatures. The technique proposed hand,The proposed technique classifies hand gestures based on the envelopes of their micro-Doppler signatures.,2034
"image color, spatial are commonly and used features The shape. texture visual layout,","The commonly used visual features are image spatial layout, color, texture and shape.",2035
"pan-tilt this for mechanism control camera paper, systems. we In an arbitrarily-assembled propose accurate","In this paper, we propose an accurate control mechanism for arbitrarily-assembled pan-tilt camera systems.",2036
"tasks. supervised in detection a training-mining used widely is weakly which framework, We object adopt","We adopt a training-mining framework, which is widely used in weakly supervised object detection tasks.",2037
"we propose AIR. limitations, these To overcome","To overcome these limitations, we propose AIR.",2038
StarCraft Strategy Time of most and popular one (SC) games. (RTS) Real is the successful,StarCraft (SC) is one of the most popular and successful Real Time Strategy (RTS) games.,2039
"player. professional between the bot top-level human large a gap the and However, remains","However, a large gap remains between the top-level bot and the professional human player.",2040
based is the on (LSTM). Our the of and the Ape-X framework combination Long-Short-Term-Memory DQN,Our framework is based on the combination of the Ape-X DQN and the Long-Short-Term-Memory (LSTM).,2041
"bot, our to as build use this named We framework LastOrder.","We use this framework to build our bot, named as LastOrder.",2042
decision form deterministic rules are the of The specified. explicitly,The form of the deterministic decision rules are explicitly specified.,2043
Likelihood shown are ratios be sufficient statistics. to,Likelihood ratios are shown to be sufficient statistics.,2044
numerical theory based testing. presented binary A prospect example is for hypothesis,A numerical example is presented for prospect theory based binary hypothesis testing.,2045
"force to or exertion getting closer the should conditions moment reality, be modeled. For systems","For getting systems closer to reality, the force or moment exertion conditions should be modeled.",2046
interpolation video propose algorithm. a frame light-weight We,We propose a light-weight video frame interpolation algorithm.,2047
a nonstationary address (BSS) source We separation problem. blind,We address a nonstationary blind source separation (BSS) problem.,2048
both The sources includes model and mixing. nonstationary,The model includes both nonstationary sources and mixing.,2049
"and an estimation joint stationarity-breaking spectra. Therefore, algorithm BSS and introduce of we deformations for","Therefore, we introduce an algorithm for joint BSS and estimation of stationarity-breaking deformations and spectra.",2050
"are performances Finally, synthetic on example. its evaluated a","Finally, its performances are evaluated on a synthetic example.",2051
still cell mechanical unknown. wall exactly is behavior How cell largely properties influence,How exactly cell wall mechanical properties influence cell behavior is still largely unknown.,2052
data publicly available is set to The the community.,The data set is publicly available to the community.,2053
for with common paper message. multicasting precoding This considers a multi-group,This paper considers precoding for multi-group multicasting with a common message.,2054
"multiple $L$ clusters, The with with communicates users. station antenna $K$ each base","The multiple antenna base station communicates with $K$ clusters, each with $L$ users.",2055
"based optimization. a alternating by suggest problem, algorithm, suboptimal on Inspired the we WMMSE","Inspired by the WMMSE problem, we suggest a suboptimal algorithm, based on alternating optimization.",2056
models years. problems learning have been used extensively recent solve in real-world to Deep,Deep learning models have been used extensively to solve real-world problems in recent years.,2057
large The relies on of performance amounts data heavily such labeled of models for training.,The performance of such models relies heavily on large amounts of labeled data for training.,2058
the segmentation. problem considering by address of video We object task this,We address this problem by considering the task of video object segmentation.,2059
existing Very few have laid of on estimation. dehazing its accurate stress methods image the,Very few of the existing image dehazing methods have laid stress on its accurate estimation.,2060
"and quantitative accurate enough. are Qualitative estimates the show, evaluations","Qualitative and quantitative evaluations show, the estimates are accurate enough.",2061
from proposed for removing A method natural new text is images.,A new method is proposed for removing text from natural images.,2062
"significantly our state-of-the-art metrics. Moreover, methods EnsNet outperform previous can of all in terms","Moreover, our EnsNet can significantly outperform previous state-of-the-art methods in terms of all metrics.",2063
possibilities of results feature show encouraging the translation. Quantitative,Quantitative results show the encouraging possibilities of feature translation.,2064
"features and among first time, reported. like For DELF the is the affinity widely-used SIFT","For the first time, the affinity among widely-used features like SIFT and DELF is reported.",2065
features. estimation hand-designed and approach tracking fail would without feature traditional visible The,The traditional hand-designed feature tracking and estimation approach would fail without visible features.,2066
"to several use are there features exploring context localization. Recently, works do to","Recently, there are several works exploring to use context features to do localization.",2067
"However, all suffer context given loss these no accuracy severe if visual of information.","However, all of these suffer severe accuracy loss if given no visual context information.",2068
address synthetic the learning domain issue effectively. from of randomized data We,We address the issue of learning from synthetic domain randomized data effectively.,2069
This be to attributed the randomization rendering distribution. parameter uniform can over,This can be attributed to uniform randomization over the rendering parameter distribution.,2070
"a projective characterization. structure-from-motion this Secondly, concrete that for utilizes algorithm propose we","Secondly, we propose a concrete algorithm for projective structure-from-motion that utilizes this characterization.",2071
be to attempt No is comprehensive. made,No attempt is made to be comprehensive.,2072
"al., these of see press). issues, some of For (Allen in further et discussion please","For further discussion of some of these issues, please see (Allen et al., in press).",2073
system. step is reading an text of scene detection important text Scene,Scene text detection is an important step of scene text reading system.,2074
"and main varied and sizes The lie aspect ratios, on arbitrary significantly shapes. orientations challenges","The main challenges lie on significantly varied sizes and aspect ratios, arbitrary orientations and shapes.",2075
"TextField generalizing to datasets. robust unseen Furthermore, in is","Furthermore, TextField is robust in generalizing to unseen datasets.",2076
"human-interpretable using most deep learning medical images. been developed applications reconstructed, However, have","However, most deep learning applications have been developed using reconstructed, human-interpretable medical images.",2077
also search random values algorithms. in We the the of investigate starting effects,We also investigate the effects of random starting values in the search algorithms.,2078
a of generating and for key models structured robotics. capable challenge Building is AI output,Building models capable of generating structured output is a key challenge for AI and robotics.,2079
cloud significant improvements methods. demonstrate We state-of-the-art against point generation,We demonstrate significant improvements against state-of-the-art point cloud generation methods.,2080
to PAs. the of enhance These will circuits overall suppression help the efficiency,These suppression circuits will help to enhance the overall efficiency of the PAs.,2081
in technology time. been a has remote hotspot a sensing detection Change for long,Change detection has been a hotspot in remote sensing technology for a long time.,2082
"are However, effective of changes usually images multi-temporal enough. methods are not complex, existing","However, changes of multi-temporal images are usually complex, existing methods are not effective enough.",2083
samples. The training unchanged high pixels pre-detection employed with to as confidence CVA find is,The CVA pre-detection is employed to find unchanged pixels with high confidence as training samples.,2084
public hyperspectral are The and two datasets a real-world on dataset. experiments performed,The experiments are performed on two real-world datasets and a public hyperspectral dataset.,2085
"estimating and one jointly. problem pose of Consequently, the becomes correspondences","Consequently, the problem becomes one of estimating pose and correspondences jointly.",2086
"a anchor we paper, for general object optimize to propose approach In detection. boxes this","In this paper, we propose a general approach to optimize anchor boxes for object detection.",2087
"anchor are adopted in Nowadays, boxes widely frameworks. state-of-the-art detection","Nowadays, anchor boxes are widely adopted in state-of-the-art detection frameworks.",2088
the success thus cervical improve screening. nuclei segmentation cancer of accurate An could rate,An accurate nuclei segmentation could thus improve the success rate of cervical cancer screening.,2089
paper of with This visual addresses generation examples. the explanations,This paper addresses the generation of explanations with visual examples.,2090
several of proposed the datasets experiments The of conducted on demonstrate method. the effectiveness results,The results of experiments conducted on several datasets demonstrate the effectiveness of the proposed method.,2091
"limited However, field. also computation has efficiency low has and it receptive","However, it has low computation efficiency and also has limited receptive field.",2092
Preliminary of model. showed results this effectiveness,Preliminary results showed effectiveness of this model.,2093
we present GAMP to Here signal for non-i.d.d. sparse estimate priors,Here we present sparse signal priors for GAMP to estimate non-i.d.d.,2094
that CSF-cNs slow We preparation. locomotion rest fictive at induced in activating confirmed the delayed,We confirmed that activating CSF-cNs at rest induced delayed slow locomotion in the fictive preparation.,2095
"of contrast, activation locomotion. promptly the In slow CSF-cNs ongoing inhibited","In contrast, the activation of CSF-cNs promptly inhibited ongoing slow locomotion.",2096
are the generative known nets to well (GANs) train. difficult remarkably It is that adversarial,It is well known that the generative adversarial nets (GANs) are remarkably difficult to train.,2097
principled towards Wasserstein GAN (WGAN) issues. creates addressing these directions recently The proposed research,The recently proposed Wasserstein GAN (WGAN) creates principled research directions towards addressing these issues.,2098
gradient training we practice in But instability. (GP-WGANs) penalty suffer WGANs found that from still,But we found in practice that gradient penalty WGANs (GP-WGANs) still suffer from training instability.,2099
control the quality. to diversity and trade-off present also between image method visual We a,We also present a method to control the trade-off between image diversity and visual quality.,2100
any burden. computation does bring It not,It does not bring any computation burden.,2101
process. neural an is formation cells emerging essentially between connections from unsupervised learning The of,The formation of connections between neural cells is emerging essentially from an unsupervised learning process.,2102
this along thalamo-cortical biological representation the model We of formation propose pathway. the a of,We propose a biological model of the formation of this representation along the thalamo-cortical pathway.,2103
domain-invariant adaptation the embedddings. distribution learning can reduce domain methods by Deep discrepancy,Deep domain adaptation methods can reduce the distribution discrepancy by learning domain-invariant embedddings.,2104
degrade the on directly classifier semantic performance dataset. can misalignment target the This,This semantic misalignment can directly degrade the classifier performance on the target dataset.,2105
"the assign target of absence images. target labels for pseudo labels, we In","In the absence of target labels, we assign pseudo labels for target images.",2106
of well two the conducted on the experiments effectiveness demonstrate datasets SCA. Extensive,Extensive experiments conducted on the two datasets well demonstrate the effectiveness of SCA.,2107
which creates pivot as is each a initialized around sphere. This shell perfect an elastic,This creates an elastic shell around each pivot which is initialized as a perfect sphere.,2108
promising accuracy achieves abdominal organ segmentation. in EBP,EBP achieves promising accuracy in abdominal organ segmentation.,2109
their appearances. human affects actions to particularly of great due videos variety and This,This particularly affects human videos due to their great variety of actions and appearances.,2110
"neural we improving study problem the networks. computational of this paper, resource utilization of In","In this paper, we study the problem of improving computational resource utilization of neural networks.",2111
"of observation This topics, a research motivates e.g. lot","This observation motivates a lot of research topics, e.g.",2112
higher computational with As models (e.g. costs,As models with higher computational costs (e.g.,2113
"To named propose method optimization a we this end, Rejuvenation. Neural novel","To this end, we propose a novel optimization method named Neural Rejuvenation.",2114
Channel been algorithms user to account to due proposed tracking channel for have mobility. variations,Channel tracking algorithms have been proposed to account for channel variations due to user mobility.,2115
synthesis. Networks face Generative realistic The Adversarial enables advance image of (GANs),The advance of Generative Adversarial Networks (GANs) enables realistic face image synthesis.,2116
of Utility implement nonparametric Models. tests to an They this application Random for test,They implement this test for an application to nonparametric tests of Random Utility Models.,2117
"paper, is challenging. their As computationally they models in such testing note","As they note in their paper, testing such models is computationally challenging.",2118
us These allow improved algorithms to tackle datasets. larger,These improved algorithms allow us to tackle larger datasets.,2119
"assumed our to inductive. is myelin model, In be","In our model, myelin is assumed to be inductive.",2120
electrode-tissue confirmed a distributed-parameter circuit. interaction the feasibility using proper findings of These studying,These findings confirmed the feasibility of studying electrode-tissue interaction using a proper distributed-parameter circuit.,2121
Both of and quantitative results demonstrate method. qualitative our suitability generality the and the,Both the qualitative and quantitative results demonstrate the generality and suitability of our method.,2122
work This in role computer the algorithms this field. of vision discusses,This work discusses the role of computer vision algorithms in this field.,2123
apply iterative algorithm an We for optimization.,We apply an iterative algorithm for optimization.,2124
data a delay receiving for introduced is Also and response. the the offloading,Also a delay is introduced for offloading the data and receiving the response.,2125
that It in also demonstrated lasting therapeutic the is tDCS applications. effect shown,It is also shown that tDCS demonstrated the lasting effect in therapeutic applications.,2126
propose motion We generative approach a to capture. physics-based,We propose a generative approach to physics-based motion capture.,2127
trajectories is from predictions of were estimated which torque active. points contact Contact and,Contact is estimated from torque trajectories and predictions of which contact points were active.,2128
statistical regressions. relationship successive numeric-computational The and simulations is obtained transfer performing by,The transfer relationship is obtained by performing successive numeric-computational simulations and statistical regressions.,2129
"a approach, DPD single energy and savings. only this complexity yielding great With is required,","With this approach, only a single DPD is required, yielding great complexity and energy savings.",2130
interval procedure bootstrap propose We a estimation. for also,We also propose a bootstrap procedure for interval estimation.,2131
progressively cascaded an framework improve phases present designed We to with autoregressive detection precision. pedestrian,We present an autoregressive pedestrian detection framework with cascaded phases designed to progressively improve precision.,2132
multimodal information. counterfactual study addresses with explanations generating This,This study addresses generating counterfactual explanations with multimodal information.,2133
"proposed we method. Moreover, information-theoretical the provide insight into","Moreover, we provide information-theoretical insight into the proposed method.",2134
types of distortion many are quality. There image which different affect,There are many different types of distortion which affect image quality.,2135
"removing on them. methods single types for of studies focused Previous distortion, have proposing","Previous studies have focused on single types of distortion, proposing methods for removing them.",2136
"image the in real to quality degrades world. due However, factors multiple","However, image quality degrades due to multiple factors in the real world.",2137
"purpose, simple architecture of effective layer a For propose networks. yet neural this we","For this purpose, we propose a simple yet effective layer architecture of neural networks.",2138
results of these the datasets. effectiveness show models Experimental real with,Experimental results show the effectiveness of these models with real datasets.,2139
"scene radiance, parameters, contents). unknown camera unknown unknown scene","unknown camera parameters, unknown scene radiance, unknown scene contents).",2140
"with tasks abundance Due of to cross-modal retrieval multiple modalities, from etc. image-text, data audio-image,","Due to abundance of data from multiple modalities, cross-modal retrieval tasks with image-text, audio-image, etc.",2141
trained The sequentially the one parts other. are network after two the of,The two parts of the network are trained sequentially one after the other.,2142
representation a a suitable This is CNN architecture. input to,This representation is a suitable input to a CNN architecture.,2143
the provided. into Interesting are resource network's distribution insights,Interesting insights into the network's resource distribution are provided.,2144
very software the is to time a Predicting for engineering software build complex managers. task,Predicting the time to build software is a very complex task for software engineering managers.,2145
of can team. the the are with productivity There development that complex directly factors interfere,There are complex factors that can directly interfere with the productivity of the development team.,2146
filling-in of task of regions inpainting the a damaged missing is incomplete or image. Image,Image inpainting is the task of filling-in missing regions of a damaged or incomplete image.,2147
in contextual It takes itself. and powerful into image account content inherent the perceptual,It takes into account powerful contextual and perceptual content inherent in the image itself.,2148
Numerical ones. compared the significantly the that outperform demonstrate models experiments proposed,Numerical experiments demonstrate that the proposed models significantly outperform the compared ones.,2149
analyze also representation hierarchical We by visualizing maps learned in network. attention generated the our,We also analyze the learned hierarchical representation by visualizing attention maps generated in our network.,2150
affected is transcranial excitability known be by stimulation. Neuronal electrical to,Neuronal excitability is known to be affected by transcranial electrical stimulation.,2151
stimulation special a design waveform frequency of We enveloped (EHFS). stimulation high,We design a special stimulation waveform of enveloped high frequency stimulation (EHFS).,2152
our the quantitatively detector dense performance over We DensePose. of improved demonstrate the pose available,We quantitatively demonstrate the improved performance of our dense pose detector over the available DensePose.,2153
pose be and detector dataset made public. will dense Our,Our dense pose dataset and detector will be made public.,2154
the unsupervised problem image-to-image We of geometric translation. address,We address the problem of unsupervised geometric image-to-image translation.,2155
"Our fashion, trained i.e. an model is unsupervised in","Our model is trained in an unsupervised fashion, i.e.",2156
training. the without during paired need of images,without the need of paired images during training.,2157
Experiments self-collected conducted large-scale datasets and on are existing of cities. both several,Experiments are conducted on both existing and self-collected large-scale datasets of several cities.,2158
We argue low-rank approximation to optimal separate is not that training. it from,We argue that it is not optimal to separate low-rank approximation from training.,2159
"integrates rank Unlike the into training. paper previous this approximation regularization works, and low","Unlike previous works, this paper integrates low rank approximation and regularization into the training.",2160
"which propose Rank and iterates low training. Pruning Trained (TRP), We rank approximation","We propose Trained Rank Pruning (TRP), which iterates low rank approximation and training.",2161
low-rank of imposes TRP the maintains constraints network capacity training. during while original,TRP maintains the capacity of original network while imposes low-rank constraints during training.,2162
framework has the a challenging proposed been learning few-shot address as setting. Meta-learning to,Meta-learning has been proposed as a framework to address the challenging few-shot learning setting.,2163
fast shows also and accuracy. study contribute ablation to both that An components convergence high,An ablation study also shows that both components contribute to fast convergence and high accuracy.,2164
data a single are unsupervised Conventional domain from adaptation (UDA) sampled training assumes domain. that,Conventional unsupervised domain adaptation (UDA) assumes that training data are sampled from a single domain.,2165
problem. three this addressing make towards contributions We major,We make three major contributions towards addressing this problem.,2166
visual to The performance. use enhance of our is driving goal work to attention autonomous,The goal of our work is to use visual attention to enhance autonomous driving performance.,2167
attention of present methods visual We two predicting maps.,We present two methods of predicting visual attention maps.,2168
analyzing images these and medical and is Capturing expensive. labor-intensive,Capturing and analyzing these medical images is labor-intensive and expensive.,2169
is proteins typically to undergo a change. Achieving require this shown task conformational to,Achieving this task is shown to typically require proteins to undergo a conformational change.,2170
scenario its implications. and A illustrates this evolutionary that model presented is physical,A physical model is presented that illustrates this evolutionary scenario and its implications.,2171
experiments be in scenario protein predictions. can implemented directly test of evolution its to The,The scenario can be implemented in experiments of protein evolution to directly test its predictions.,2172
"results, identification interpretation. recommend provide new We discuss estimate, parameters several and to their formal","We provide formal identification results, recommend several new parameters to estimate, and discuss their interpretation.",2173
"out methods for we carrying Further, propose inference.","Further, we propose methods for carrying out inference.",2174
ambient of for living and Detecting alone. unintended essential people intelligence elderly is healthcare falls,Detecting unintended falls is essential for ambient intelligence and healthcare of elderly people living alone.,2175
cameras that the Event gather temporal sensors are of scene. biologically-inspired evolution the,Event cameras are biologically-inspired sensors that gather the temporal evolution of the scene.,2176
stream variations brightness capture They of a pixel-wise events. output corresponding asynchronous and,They capture pixel-wise brightness variations and output a corresponding stream of asynchronous events.,2177
"we present scenario discovery an In adaptive counting. framework crowd paper, this for","In this paper, we present an adaptive scenario discovery framework for crowd counting.",2178
"can and from estimated be development data. general, satisfactorily literature starting In fecundity","In general, development and fecundity can satisfactorily be estimated starting from literature data.",2179
assigning their clears It or issue responsibility the to machines of up creators.,It clears up the issue of assigning responsibility to machines or their creators.,2180
"network. a stream In is DST-GCNN two particular,","In particular, DST-GCNN is a two stream network.",2181
several layers to are such future stacked Then flows time. predict over together,Then several such layers are stacked together to predict future flows over time.,2182
saliency non-discriminative. maps produced The are proven to be them by,The saliency maps produced by them are proven to be non-discriminative.,2183
explanations LRP the analysis not by class-discriminative. conclude and The experiments generated that are,The experiments and analysis conclude that the explanations generated by LRP are not class-discriminative.,2184
the an explanations We evaluate Game quantitatively Pointing also ablation a and with study.,We also evaluate the explanations quantitatively with a Pointing Game and an ablation study.,2185
quantitative that Both explanations better the show generates the CLRP and qualitative than LRP. evaluations,Both qualitative and quantitative evaluations show that the CLRP generates better explanations than the LRP.,2186
efficiently regional object requires among instances image cluttered representations. comprehensive Retrieving compact scenes yet,Retrieving object instances among cluttered scenes efficiently requires compact yet comprehensive regional image representations.,2187
"focuses Intuitively, index regions. semantics can build object the on help that the most relevant","Intuitively, object semantics can help build the index that focuses on the most relevant regions.",2188
"this provided. general ARCH on a paper, spatial and In models overview is spatiotemporal","In this paper, a general overview on spatial and spatiotemporal ARCH models is provided.",2189
"ARCH-type we different particular, spatial between three In models. distinguish","In particular, we distinguish between three different spatial ARCH-type models.",2190
definition addition al. et to original Otto of the In,In addition to the original definition of Otto et al.,2191
"parameters are the For maximum-likelihood estimators this for proposed. model, new","For this new model, maximum-likelihood estimators for the parameters are proposed.",2192
"of spatial we consider ARCH the addition, a definition new In process. complex-valued","In addition, we consider a new complex-valued definition of the spatial ARCH process.",2193
"R-package of use the of practical demonstrated. is the From point view, spGARCH a","From a practical point of view, the use of the R-package spGARCH is demonstrated.",2194
"Eventually, all apply to we a procedures example. real-data","Eventually, we apply all procedures to a real-data example.",2195
less makes to training sensitive models circuit CNN The fine-tuned the uncertainty. the,The fine-tuned training makes the CNN models less sensitive to the circuit uncertainty.,2196
apply faulty method CNN and design to an the accelerator. a both accelerator We overclocked,We apply the design method to both an overclocked CNN accelerator and a faulty accelerator.,2197
"cognition methods insufficient fallacious and create understanding. because However, of textures learning-based often artifacts existing","However, existing learning-based methods often create artifacts and fallacious textures because of insufficient cognition understanding.",2198
Human constant target the of is attribute. regardless cognition,Human cognition is constant regardless of the target attribute.,2199
more comparisons previous ours completions. that obtain state-of-the-art much with natural methods image Quantitative show,Quantitative comparisons with previous state-of-the-art methods show that ours obtain much more natural image completions.,2200
and editing an parsing numerous model scene insertion image applications. potentially facilitate Such can object,Such an object insertion model can potentially facilitate numerous image editing and scene parsing applications.,2201
network and are modules via two a The together trained. connected transformation jointly spatial,The two modules are connected together via a spatial transformation network and jointly trained.,2202
Medication one can which Events administration occur. Drug pathway Adverse is by (ADE),Medication administration is one pathway by which Adverse Drug Events (ADE) can occur.,2203
in in-the-wild and images problem. an challenging is important Facial captured landmark localisation,Facial landmark localisation in images captured in-the-wild is an important and challenging problem.,2204
"this propose In U-Nets we dense work, stacked task. this for innovatively","In this work, we innovatively propose stacked dense U-Nets for this task.",2205
up. datasets are Such hard scale to thus,Such datasets are thus hard to scale up.,2206
discriminators. synthetic context image refined The with is adversarial into and two object the blended,The synthetic object is refined and blended into the image context with two adversarial discriminators.,2207
newly The on (t-SVD) singular directions. value light new proposed decomposition tensor sheds,The newly proposed tensor singular value decomposition (t-SVD) sheds light on new directions.,2208
the direction the alternating of method under framework. derive further We multipliers algorithm,We further derive the algorithm under the alternating direction method of multipliers framework.,2209
several addresses local ambiguity method challenges LSD and rigorously in class imbalance. Our as such,Our method rigorously addresses several challenges in LSD such as local ambiguity and class imbalance.,2210
independent an and identically (i.i.d.) distributed Assuming,Assuming an independent and identically distributed (i.i.d.),2211
dictionaries. Our applies or discrete continuous on procedure both,Our procedure applies on both discrete or continuous dictionaries.,2212
"by by bidirectional models. decoded Then, are and task-specific visual TreeLSTM encoded contexts","Then, visual contexts are encoded by bidirectional TreeLSTM and decoded by task-specific models.",2213
the results experimental methods. show method outperforms Extensive state-of-the-art several that proposed,Extensive experimental results show that the proposed method outperforms several state-of-the-art methods.,2214
on and results real We qualitative videos. world present quantitative,We present quantitative and qualitative results on real world videos.,2215
"assessment, would imaging reliable in-vivo For be and real-time immediate required.","For immediate assessment, reliable and real-time in-vivo imaging would be required.",2216
"human So by the far, is done analysis experts.","So far, the analysis is done by human experts.",2217
small learning very architectures. state-of-the-art We transfer with overcome sizes through dataset,We overcome very small dataset sizes through transfer learning with state-of-the-art architectures.,2218
"based paper algorithm inference efficient This object describes for deep-learning AutoFocus, detectors. an multi-scale","This paper describes AutoFocus, an efficient multi-scale inference algorithm for deep-learning based object detectors.",2219
"The processing detector only applied computation FocusChips, inside which reduces is scales. finer while","The detector is only applied inside FocusChips, which reduces computation while processing finer scales.",2220
"Blocksworld report, In this for Photo-realistic an we generator artificial dataset introduce domain.","In this report, we introduce an artificial dataset generator for Photo-realistic Blocksworld domain.",2221
"when applied natural problem adaptation domain images. this to However, in results","However, this results in domain adaptation problem when applied to natural images.",2222
"training. propose over to framework paper, this In challenges a using we these improve adversarial","In this paper, we propose a framework to improve over these challenges using adversarial training.",2223
the of architecture. simpler achieve art We results with competitive a the much state with,We achieve results competitive with the state of the art with a much simpler architecture.,2224
obtain rotation to down pedestrian We viewpoint problem the focused narrow conclusions. the angle to,We narrow the viewpoint problem down to the pedestrian rotation angle to obtain focused conclusions.,2225
"regard, the community. In this to makes this contributions two paper","In this regard, this paper makes two contributions to the community.",2226
"introduce large-scale synthetic we a data First, engine, PersonX.","First, we introduce a large-scale synthetic data engine, PersonX.",2227
in deeper Extensive experiment a the helps problems have fundamental re-ID. understanding of us person,Extensive experiment helps us have a deeper understanding of the fundamental problems in person re-ID.,2228
to inverse approach problem. The is radar applied imaging proposed aperture synthetic the (ISAR),The proposed approach is applied to the inverse synthetic aperture radar (ISAR) imaging problem.,2229
"COCO on segmentation interactive to we compare panoptic single dataset. the object Finally,","Finally, we compare to interactive single object segmentation on the COCO panoptic dataset.",2230
"inference inductive contextual in to use the bias and collocations we humans Intuitively, discourse. compose","Intuitively, we humans use the inductive bias to compose collocations and contextual inference in discourse.",2231
involving have spectrum images. Many and biometrics infra-red both (VIS) visible (IR) researchers explored periocular,Many researchers have explored periocular biometrics involving both visible (VIS) and infra-red (IR) spectrum images.,2232
method space. neutrosophic choroidal of segmentation an paper presents This automatic images from in EDI-OCT,This paper presents an automatic method of choroidal segmentation from EDI-OCT images in neutrosophic space.,2233
"same (CNNs) convolutional the process networks filters. of with different Conventionally, neural set images","Conventionally, convolutional neural networks (CNNs) process different images with the same set of filters.",2234
"challenge in this images a However, pose variations fashion. to the","However, the variations in images pose a challenge to this fashion.",2235
"obtain sample-specific intermediate maps to from the we features, In order autoencoder. extract feature an","In order to obtain sample-specific features, we extract the intermediate feature maps from an autoencoder.",2236
of application an proposed the desirable estimators. studies illustrate the performance finite-sample Simulation empirical and,Simulation studies and an empirical application illustrate the desirable finite-sample performance of the proposed estimators.,2237
Open-source proposed the is tools evaluation software implementing available. for policy,Open-source software for implementing the proposed policy evaluation tools is available.,2238
tools making. are in widely pest control models used Phenological decision,Phenological models are widely used tools in pest control decision making.,2239
models only. built are function developmental often considering the Phenological rate,Phenological models are often built considering the developmental rate function only.,2240
The mainly to generate of reason the flow difficulty lies in annotation ground-truth. optical human,The reason lies mainly in the difficulty of human annotation to generate optical flow ground-truth.,2241
the several propose to optical flow We ways augment variations.,We propose several ways to augment the optical flow variations.,2242
focuses policies demonstration. decentralized sub-optimal for paper tasks on This learning such using,This paper focuses on learning decentralized policies for such tasks using sub-optimal demonstration.,2243
approaches. outperforms state-of-the-art many based It demonstration significantly,It significantly outperforms many state-of-the-art demonstration based approaches.,2244
complex This to leads more road traffic management.,This leads to more complex road traffic management.,2245
"microscopic a SUMO, traffic system simulator. is executed suggested The and well-known using simulated","The suggested system is executed and simulated using SUMO, a well-known microscopic traffic simulator.",2246
"system. results obtained efficiency the encouraging, proposed confirming the are The of","The obtained results are encouraging, confirming the efficiency of the proposed system.",2247
point is Processing a large clouds task. challenging,Processing large point clouds is a challenging task.,2248
"is often processed that to easily. the sampled more a can size Therefore, data be","Therefore, the data is often sampled to a size that can be processed more easily.",2249
sample data? The to is the question how,The question is how to sample the data?,2250
is A Point Sampling technique Farthest popular sampling (FPS).,A popular sampling technique is Farthest Point Sampling (FPS).,2251
"agnostic downstream a to application is However, etc. FPS retrieval, (classification,","However, FPS is agnostic to a downstream application (classification, retrieval, etc.",2252
is sample. We it to better how show to learn that,We show that it is better to learn how to sample.,2253
"original subset step. Therefore, in a post-processing points match a it to of the we","Therefore, we match it to a subset of the original points in a post-processing step.",2254
We problem or clipped of the sparse measurements. quantized recovering from a signal address,We address the problem of recovering a sparse signal from clipped or quantized measurements.,2255
also are A brief and UAV-to-UAV of hardware discussion provided. communication issues,A brief discussion of UAV-to-UAV communication and hardware issues are also provided.,2256
applications computer vision. model has fields motion various prediction Human in of,Human motion prediction model has applications in various fields of computer vision.,2257
"training, further regularize strategy. a the novel introduce prediction we recursive To","To further regularize the training, we introduce a novel recursive prediction strategy.",2258
"MSI denoising. (LTDL) Dictionary method for In Low-rank this a we Tensor propose Learning paper,","In this paper, we propose a Low-rank Tensor Dictionary Learning (LTDL) method for MSI denoising.",2259
"blocks from we them Firstly, MSI cluster extract and groups. into the","Firstly, we extract blocks from the MSI and cluster them into groups.",2260
dictionary of validate the by the on Experiments learning LTDL. data effectiveness synthetic,Experiments on synthetic data validate the effectiveness of dictionary learning by the LTDL.,2261
many crucial for video applications. Object is detection in,Object detection in video is crucial for many applications.,2262
"problem. images, Compared the can help video additional to which detection disambiguate provides cues to","Compared to images, video provides additional cues which can help to disambiguate the detection problem.",2263
"To tubes we temporal to bounding space-time temporal boxes. model corresponding evolution, sequences of introduce","To model temporal evolution, we introduce space-time tubes corresponding to temporal sequences of bounding boxes.",2264
"We tubes, architectures classifying for propose respectively. generating two CNN and","We propose two CNN architectures for generating and classifying tubes, respectively.",2265
implements the then video. The Tube-CNN a object in tube-level detector,The Tube-CNN then implements a tube-level object detector in the video.,2266
in models scenes. particular difficult dynamic advantages show Tube,Tube models show particular advantages in difficult dynamic scenes.,2267
the temporal introduce enforce regularity We in to motion predicted first-order sequence. loss explicitly gradient,We introduce first-order gradient loss to explicitly enforce temporal regularity in the predicted motion sequence.,2268
hardware. new approach quantum This a computing novel to paper uses that image presents segmentation,This paper presents a novel approach to image segmentation that uses new quantum computing hardware.,2269
and current algorithm computers. This implemented be can on quantum near-term,This algorithm can be implemented on current and near-term quantum computers.,2270
and are data. Encouraging medical imaging on presented results artificial,Encouraging results are presented on artificial and medical imaging data.,2271
"an This vision. leveraging practical towards computer represents step computers quantum for important,","This represents an important, practical step towards leveraging quantum computers for computer vision.",2272
"interesting. we may situation and more different, complex arguably the Here be that show more","Here we show that the situation may be different, more complex and arguably more interesting.",2273
that principle coexist. Competitive Complete the cannot Gauze We competitors that exclusion well-known discuss states,We discuss the well-known Gauze Competitive exclusion principle that states that Complete competitors cannot coexist.,2274
(UE). photodiodes placing two consider on user (PDs) the configurations for the equipment We,We consider two configurations for placing the photodiodes (PDs) on the user equipment (UE).,2275
"model. we the For the (ORWP) random mobility orientation-based waypoint invoke latter,","For the latter, we invoke the orientation-based random waypoint (ORWP) mobility model.",2276
energy is highly (MDT) transmitter It with adaptive multi-directional is the shown that SM efficient.,It is shown that the multi-directional transmitter (MDT) with adaptive SM is highly energy efficient.,2277
object the image. grounded and the in,and the object grounded in the image.,2278
a leisure for The (PJ) jigsaw is game humans. well-known pictorial puzzle,The pictorial jigsaw (PJ) puzzle is a well-known leisure game for humans.,2279
to We intelligence. PJ Internet-based by approach an human collective puzzle propose stigmergy-inspired solving,We propose an approach to solving PJ puzzle by stigmergy-inspired Internet-based human collective intelligence.,2280
"we Network present the (KTN). In Transformer work, Kernel this","In this work, we present the Kernel Transformer Network (KTN).",2281
the recognition This multiple application re-training enables to KTN. tasks without,This enables application to multiple recognition tasks without re-training the KTN.,2282
"cooperate cancer for growth example, producing factors. In by tumour cells can","In cancer for example, tumour cells can cooperate by producing growth factors.",2283
attend Most once captioning attention-based to per the word. image image models,Most attention-based image captioning models attend to the image once per word.,2284
"miss easy once is to is attending However, rigid some per word information. and","However, attending once per word is rigid and is easy to miss some information.",2285
outperforms Experiments dataset caption that our state-of-the-art. show on the MSCOCO method the,Experiments on the MSCOCO caption dataset show that our method outperforms the state-of-the-art.,2286
This image problem super-resolution rise the to document of (BDISR). gives binary,This gives rise to the problem of binary document image super-resolution (BDISR).,2287
We analyze address their architectures propose multiple problem and network neural this to deep performance.,We propose multiple deep neural network architectures to address this problem and analyze their performance.,2288
mmWave combiners a and Configuring to precoders systems. hybrid practical deploy is challenge communication major,Configuring hybrid precoders and combiners is a major challenge to deploy practical mmWave communication systems.,2289
study compression scheme periodic present We and lossless signals. for of causal real-valued a,We present and study a scheme for lossless causal compression of periodic real-valued signals.,2290
such utilize semantic understanding. generative achieve networks Current adversarial (GANs) image to inpainting approaches,Current image inpainting approaches utilize generative adversarial networks (GANs) to achieve such semantic understanding.,2291
"algorithm. propose we a sequential problem, SeFM feature point matching overcome this To","To overcome this problem, we propose SeFM a sequential feature point matching algorithm.",2292
epipole of each the to We find image. utilize epipolar geometry the first,We first utilize the epipolar geometry to find the epipole of each image.,2293
"wrong validity. matching calculating also by the Furthermore, dot-pairs we remove","Furthermore, we also remove wrong matching dot-pairs by calculating the validity.",2294
"recall. SIFT Moreover, it and SURF, such as outperforms conventional regarding precision and algorithms,","Moreover, it outperforms conventional algorithms, such as SIFT and SURF, regarding precision and recall.",2295
"these carefully compelling results. crafted, can When regularizations produce","When carefully crafted, these regularizations can produce compelling results.",2296
Human-Engineered amalgamation Features of (ALF) an Auto-Learned is and HDL (HEF). Features,HDL is an amalgamation of Auto-Learned Features (ALF) and Human-Engineered Features (HEF).,2297
"measurement effective Radiological and offers disease in diagnosis assessment. is of which useful imaging anatomy,","Radiological imaging offers effective measurement of anatomy, which is useful in disease diagnosis and assessment.",2298
"of segmentation the structures from the Nevertheless, time-consuming. very is medical still images atrial left","Nevertheless, the segmentation of the left atrial structures from medical images is still very time-consuming.",2299
"an ensemble to Finally, models we segment integrate prediction in datasets. trained testing several","Finally, we integrate several trained models in an ensemble prediction to segment testing datasets.",2300
as are estimates multidimensional points. The stored,The estimates are stored as multidimensional points.,2301
is Initial promising. that proposed indicate the the of method performance results,Initial results indicate that the performance of the proposed method is promising.,2302
Transformer human introduce video recognizing localizing in clips. and We actions the model Action for,We introduce the Action Transformer model for recognizing and localizing human actions in video clips.,2303
object discovery of problem the We dense for videos. segmentation consider masks in providing,We consider the problem of providing dense segmentation masks for object discovery in videos.,2304
$t$-based construct to intervals. use can confidence values $k$ these We,We can use these $k$ values to construct $t$-based confidence intervals.,2305
"a called We propose task, Story Visualization. new","We propose a new task, called Story Visualization.",2306
"evaluate the datasets. To the CLEVR-SV datasets existing and to Pororo-SV model, modified create we","To evaluate the model, we modified existing datasets to create the CLEVR-SV and Pororo-SV datasets.",2307
"and evaluation. contextual metrics, consistency Empirically, models human image quality, state-of-the-art outperforms StoryGAN in","Empirically, StoryGAN outperforms state-of-the-art models in image quality, contextual consistency metrics, and human evaluation.",2308
"proposed. methods observation, upon Based the many attack were","Based upon the observation, many attack methods were proposed.",2309
in method our prove the and intuitively empirically space. search reducing the rationality of We,We intuitively and empirically prove the rationality of our method in reducing the search space.,2310
is Numerical obtained demonstrate over significant gain results methods. performance conventional that,Numerical results demonstrate that significant performance gain is obtained over conventional methods.,2311
"tasks. advantages computation-intensive Also, more are handling latency-critical the when proposed algorithm prominent the of","Also, the advantages of the proposed algorithm are more prominent when handling computation-intensive latency-critical tasks.",2312
energy Civil to changing of dissipation structures decline. capacity which leads are the on verge,Civil structures are on the verge of changing which leads energy dissipation capacity to decline.,2313
"Mode EEMD more Empirical Albeit (IMFs) Decomposition than Functions brings (EMD). appropriate Mode analogous, Intrinsic","Albeit analogous, EEMD brings more appropriate Intrinsic Mode Functions (IMFs) than Empirical Mode Decomposition (EMD).",2314
frequency mode mode to assess shape. and are IMFs first employed,IMFs are employed to assess first mode frequency and mode shape.,2315
"predicted measured acceleration damage. onset and congruency between Therefore, provides any of","Therefore, any congruency between predicted and measured acceleration provides onset of damage.",2316
estimate to method Then ANN stiffness matrix. another applied is,Then another ANN method is applied to estimate stiffness matrix.,2317
This one and the implemented steel moment-resisting are is nonlinear acceptable. on results frame algorithm,This algorithm is implemented on one nonlinear moment-resisting steel frame and the results are acceptable.,2318
the biometric widely Fingerprints identification. form are deployed most of,Fingerprints are the most widely deployed form of biometric identification.,2319
same share the have identifiers. fingerprint unique because two No individuals biometric they,No two individuals share the same fingerprint because they have unique biometric identifiers.,2320
paper This fingerprint presents accuracy. improves verification algorithm which efficient an matching,This paper presents an efficient fingerprint verification algorithm which improves matching accuracy.,2321
get and Fingerprint skin and due in to impression images degraded conditions. variations corrupted,Fingerprint images get degraded and corrupted due to variations in skin and impression conditions.,2322
"enhancement detection singular minutiae techniques and employed to Thus, point are prior image extraction.","Thus, image enhancement techniques are employed prior to singular point detection and minutiae extraction.",2323
is curvature. point Singular the point maximum of,Singular point is the point of maximum curvature.,2324
to the particular on depending radius The the accuracy according application. varied can be,The radius can be varied according to the accuracy depending on the particular application.,2325
that more effective the proposed robust. measures is algorithm The and demonstrate,The measures demonstrate that the proposed algorithm is more effective and robust.,2326
"to the vulnerable consequence, large can techniques As a occlusions. resulting be","As a consequence, the resulting techniques can be vulnerable to large occlusions.",2327
"Furthermore, relies real-time achieve architecture on performance. a to simple it enough","Furthermore, it relies on a simple enough architecture to achieve real-time performance.",2328
networks known attacks. adversarial neural be to are vulnerable to Deep,Deep neural networks are known to be vulnerable to adversarial attacks.,2329
exposes in and of security-sensitive exploits to their them highlights This robustness. applications potential lack,This exposes them to potential exploits in security-sensitive applications and highlights their lack of robustness.,2330
and It potential for improvement has more also training. flexibility via,It also has more flexibility and potential for improvement via training.,2331
proposed staged used the supervision of several in is outputs Intermediate architecture approach. a in,Intermediate supervision is used in several outputs of the proposed architecture in a staged approach.,2332
"This is commonly to as referred ""word spotting"".","This is commonly referred to as ""word spotting"".",2333
further introduce called version a Ctrl-F-Mini. simplified We,We further introduce a simplified version called Ctrl-F-Mini.,2334
"with is segmented similar though faster to limited manuscripts. performance, easily more It is it","It is faster with similar performance, though it is limited to more easily segmented manuscripts.",2335
these problem. of numerical a optimization applications Many large-scale involve solving,Many of these applications involve solving a large-scale numerical optimization problem.,2336
"instance interactive segmentation, to feedback segmentation masks. give refine iteratively users In","In interactive instance segmentation, users give feedback to iteratively refine segmentation masks.",2337
visual which is equally are in multimodal essential an and auditory important. It contents task,It is an essential multimodal task in which auditory and visual contents are equally important.,2338
How sentences words are the with modalities? auditory the the generated in visual and associated,How are the words in the generated sentences associated with the auditory and visual modalities?,2339
"benchmarks. several state-of-the-arts consistently Overall, on the NMTree proposed the outperforms","Overall, the proposed NMTree consistently outperforms the state-of-the-arts on several benchmarks.",2340
grounding score calculation in great results detail. show Qualitative explainable,Qualitative results show explainable grounding score calculation in great detail.,2341
most in artworks the Chinese of painting one traditional world. historical is the,Chinese traditional painting is one of the most historical artworks in the world.,2342
to popular Asia being Southeast due aesthetically It in very appealing. Eastern and is,It is very popular in Eastern and Southeast Asia due to being aesthetically appealing.,2343
"with painting, artistic it western textureless. usually and visually abstract is more Compared","Compared with western artistic painting, it is usually more visually abstract and textureless.",2344
neural It style abstraction problem transfer. in remains to challenging preserve a,It remains a challenging problem to preserve abstraction in neural style transfer.,2345
"this Transfer paper, Neural we Abstract method for traditional a present Style Chinese painting. In","In this paper, we present a Neural Abstract Style Transfer method for Chinese traditional painting.",2346
"recognition. problem the sketch of paper, address In hand-drawn we this","In this paper, we address the problem of hand-drawn sketch recognition.",2347
"observe fraction a However, this small very of only cameras distribution. most","However, most cameras observe only a very small fraction of this distribution.",2348
Recognition active an has the in become domains research recently topic community. across,Recognition across domains has recently become an active topic in the research community.,2349
"However, domains. it recognition has new the unseen largely of problem in in overlooked been","However, it has been largely overlooked in the problem of recognition in new unseen domains.",2350
"we benchmark recognition, digit On four on i.e. recognition popular digit databases,","On digit recognition, we benchmark on four popular digit recognition databases, i.e.",2351
accepted software principle tests has a is bugs. It without widely that,It is a widely accepted principle that software without tests has bugs.,2352
"robust populations. in The developmental, analysis is patient healthy, automated and and fully","The analysis is fully automated and robust in healthy, developmental, and patient populations.",2353
challenging state-of-the-art datasets. localization against methods evaluate three on (SOTA) We our model,We evaluate our model against state-of-the-art (SOTA) methods on three challenging localization datasets.,2354
maps We insights offer show into model. to predicted attention the finally our useful,We finally show the predicted attention maps to offer useful insights into our model.,2355
"massively within undersampled mechanisms. about measurements us Therefore, tell little correlations","Therefore, correlations within massively undersampled measurements tell us little about mechanisms.",2356
different are challenges and for ways the tackle both too. to vary The them,The challenges for both are different and the ways to tackle them vary too.,2357
easier makes learning showed and also things deep precise. We,We also showed deep learning makes things easier and precise.,2358
"a alignment is challenging problem Therefore, VSR. yet important for temporal","Therefore, temporal alignment is a challenging yet important problem for VSR.",2359
"to supporting TDAN frames kernels, frame. align the reference using with transforms the corresponding By","By using the corresponding kernels, TDAN transforms supporting frames to align with the reference frame.",2360
of the effectiveness the Experimental model. results proposed VSR TDAN-based demonstrate,Experimental results demonstrate the effectiveness of the proposed TDAN-based VSR model.,2361
"star sensors. conventional most Currently, optical use trackers","Currently, most star trackers use conventional optical sensors.",2362
"this propose star for usage the application we event paper, tracking. In of sensors","In this application paper, we propose the usage of event sensors for star tracking.",2363
"community, familiar researchers. remains research many topic the less to Within AI this","Within the AI research community, this topic remains less familiar to many researchers.",2364
surveillance. camera role important video person an a Finding plays a in across network,Finding a person across a camera network plays an important role in video surveillance.,2365
of spatial-temporal person current (ReID) a constraint. Most neglect re-identification methods,Most of current person re-identification (ReID) methods neglect a spatial-temporal constraint.,2366
"a Histogram-Parzen To develop probability fast distribution, method. approximate spatial-temporal (HP) a we complex","To approximate a complex spatial-temporal probability distribution, we develop a fast Histogram-Parzen (HP) method.",2367
"from performance on manipulation unseen significantly. However, drops examples their approaches,","However, on examples from unseen manipulation approaches, their performance drops significantly.",2368
"transferability, Forensic-Transfer To this introduce limitation in we (FT). address","To address this limitation in transferability, we introduce Forensic-Transfer (FT).",2369
named Scale Peak-to-Correlation SITUP: system resulting The using is Tracking Energy. Average Invariant,The resulting system is named SITUP: Scale Invariant Tracking using Average Peak-to-Correlation Energy.,2370
filters are in Filters enhancing medical used one Frangi widely vessels the of for images.,Frangi Filters are one of the widely used filters for enhancing vessels in medical images.,2371
"for enhancing manually coronary fluoroscope, images. for are angiogram thresholds These changed individual","These thresholds are changed manually for individual fluoroscope, for enhancing coronary angiogram images.",2372
not best the has our of been addressed problem This yet. to knowledge,This problem to the best of our knowledge has not been addressed yet.,2373
estimation (DOA) considered. using of direction-of-arrival arrays The problem non-uniform is,The problem of direction-of-arrival (DOA) estimation using non-uniform arrays is considered.,2374
"with a perform In is sparsity-based interpolation DOA presented to increased technique estimation addition, DOFs.","In addition, a sparsity-based interpolation technique is presented to perform DOA estimation with increased DOFs.",2375
ideal operational aforementioned approaches with The scenarios. deal,The aforementioned approaches deal with ideal operational scenarios.,2376
are included. methods simulations numerical validate also Extensive which different proposed the,Extensive numerical simulations which validate the different proposed methods are also included.,2377
We for verifying fairness specifications. a scalable design algorithm,We design a scalable algorithm for verifying fairness specifications.,2378
object without could for detection applications immense real-world methodology unlock Our datasets.,Our methodology could unlock object detection for real-world applications without immense datasets.,2379
"facial of application examine the we In particular, recognition.","In particular, we examine the application of facial recognition.",2380
"lines, for these weight a Along via we backdooring describe methodology general perturbations. targeted CNNs","Along these lines, we describe a general methodology for backdooring CNNs via targeted weight perturbations.",2381
depth monocular present framework single unsupervised videos. view estimation a using novel learning for We,We present a novel unsupervised learning framework for single view depth estimation using monocular videos.,2382
may result overfitting. in baseline short This and a,This may result in a short baseline and overfitting.,2383
"impute resolution low image. generally a works low resolution a Besides, estimate from previous depth","Besides, previous works generally estimate a low resolution depth from a low resolution impute image.",2384
is then depth resolution. low to The interpolated original the resolution recover,The low resolution depth is then interpolated to recover the original resolution.,2385
"we clip to deal Additionally, occlusion. loss with introduce and the moving objects","Additionally, we introduce the clip loss to deal with moving objects and occlusion.",2386
for blood regional myocardial available and flow were Manual segmentation (MBF) comparison.,Manual segmentation and regional myocardial blood flow (MBF) were available for comparison.,2387
"decreased rate positive false the and systematically rate false $\beta$ systematically As increased. negative increased,","As $\beta$ increased, the false positive rate systematically decreased and false negative rate systematically increased.",2388
two for uncertainty. simple model introduce also methods We assessing,We also introduce two simple methods for assessing model uncertainty.,2389
pose end-to-end previous an performed mainly architecture refinement The in multi-stage was methods. trainable through,The pose refinement was performed mainly through an end-to-end trainable multi-stage architecture in previous methods.,2390
"pose careful and However, require estimation design. models on dependent are they model highly","However, they are highly dependent on pose estimation models and require careful model design.",2391
"propose model-agnostic contrast, method. refinement we pose a By","By contrast, we propose a model-agnostic pose refinement method.",2392
happening a area. Interreflections its surface concave to lead of values over gradients RGB in,Interreflections happening in a concave surface lead to gradients of RGB values over its area.,2393
only interreflection trained simulated model. physics-based Our data constructed with using is a network,Our network is trained with only simulated data constructed using a physics-based interreflection model.,2394
"to approaches. addition, on compared data interreflection-based better In results gives other real it","In addition, it gives better results on real data compared to other interreflection-based approaches.",2395
"and Then lesion-targeted classifiers, compare respectively. we train both and disease-targeted","Then we train and compare both disease-targeted and lesion-targeted classifiers, respectively.",2396
often a process. localization The part as arises of of problem navigation,The problem of localization often arises as part of a navigation process.,2397
"input contrast, train we joint and classification data. of reconstruction networks for In","In contrast, we train networks for joint classification and reconstruction of input data.",2398
are assembly. drivers tradeoffs of Life-history among species community major,Life-history tradeoffs among species are major drivers of community assembly.,2399
tradeoffs investigate how promote of studies species. deterministic Most coexistence,Most studies investigate how tradeoffs promote deterministic coexistence of species.,2400
"machine garment cost. is desirable to sewing automatic an manufacturing, In reduce","In garment manufacturing, an automatic sewing machine is desirable to reduce cost.",2401
and framework blob motion for matching. using Our includes feature a detection module estimation,Our framework includes a module for motion estimation using blob detection and feature matching.,2402
pattern facilitate It also module the lattice weave includes a detection for recognition. to,It also includes a module for lattice detection to facilitate the weave pattern recognition.,2403
video networks present for We recognition. SlowFast,We present SlowFast networks for video recognition.,2404
"We major accuracy state-of-the-art video report AVA. Charades and recognition on Kinetics, benchmarks,","We report state-of-the-art accuracy on major video recognition benchmarks, Kinetics, Charades and AVA.",2405
"evaluate our datasets, on three namely approach MMI. Oulu-CASIA, We and CK+ popular","We evaluate our approach on three popular datasets, namely Oulu-CASIA, CK+ and MMI.",2406
"made ray to a the ray through is data. pass casting, In light of volume","In ray casting, a ray of light is made to pass through the volume data.",2407
"demonstrate tend on real performance to However, images. trained synthetic models simply poor on images","However, models trained simply on synthetic images tend to demonstrate poor performance on real images.",2408
One approach inputs. a as that is depth uses fusion-based images,One is a fusion-based approach that uses depth images as inputs.,2409
learning The a other images approach that outputs. multitask uses is as depth,The other is a multitask learning approach that uses depth images as outputs.,2410
missing in a plane. time-frequency reconstruction problem data We new propose the of,We propose a new problem of missing data reconstruction in the time-frequency plane.,2411
A mathematical problem given. this is of formulation,A mathematical formulation of this problem is given.,2412
We algorithms. existing propose alternatives of three,We propose three alternatives of existing algorithms.,2413
and optimization algorithms: two PhaseLift An PhaseCut. and semidefinite Griffin programming algorithm: iterative Lim and,An iterative algorithm: Griffin and Lim and two semidefinite programming optimization algorithms: PhaseLift and PhaseCut.,2414
knowledge results quality. of certain phases that obtained The reconstruction's show improves the,The obtained results show that knowledge of certain phases improves the reconstruction's quality.,2415
task. can very time-consuming a Identifying them be,Identifying them can be a very time-consuming task.,2416
by problem. it perturbation data (image) accomplished modeling is as This,This is accomplished by modeling it as data (image) perturbation problem.,2417
is Unconstrained stimulating text recognition. in field pattern recognition the a of branch,Unconstrained text recognition is a stimulating field in the branch of pattern recognition.,2418
Hidden modeling. are Models Markov used then for The sequences,The Hidden Markov Models are then used for sequences modeling.,2419
method promising datasets. both on recognition Our achieves,Our method achieves promising recognition on both datasets.,2420
neural of computing the computational such reducing costs inference. require Important network as applications mobile,Important applications such as mobile computing require reducing the computational costs of neural network inference.,2421
single settings. a propose framework efficiency both end-to-end improve in inference can that We,We propose a single end-to-end framework that can improve inference efficiency in both settings.,2422
"end-to-end, We pruning as as computation. well supports same and train conditional the technique","We train end-to-end, and the same technique supports pruning as well as conditional computation.",2423
is Automatic vertebrae localization images identification challenging. CT arbitrary and from,Automatic vertebrae identification and localization from arbitrary CT images is challenging.,2424
have space efficient heuristic tested. search regard methods shortest-path Five with been to,Five heuristic methods with regard to efficient shortest-path search space have been tested.,2425
very a topic has mapping recent Cross-domain in active been years.,Cross-domain mapping has been a very active topic in recent years.,2426
"state-of-the-art i.e. two Most of on the focused has existing reasons, only","Most of the existing state-of-the-art has focused only on two reasons, i.e.",2427
"on and (i) and either (iii). or (i) (ii),","either on (i) and (ii), or (i) and (iii).",2428
"all obviously this While attractive it approach, applicable is not in an scenarios. is","While this is an obviously attractive approach, it is not applicable in all scenarios.",2429
communication (CSI) any crucial information for Channel very state is wireless systems.,Channel state information (CSI) is very crucial for any wireless communication systems.,2430
"at be characterized can side Typically, impulse CSI using receiver the channel response (CIR).","Typically, CSI can be characterized at the receiver side using channel impulse response (CIR).",2431
shown wireless the observations have often sparse. Many of that multipath are broadband channels CIR,Many observations have shown that the CIR of broadband multipath wireless channels are often sparse.,2432
"model/type. category, one humans always vehicle's coarse-grained firstly the i.e., Specifically, determine car","Specifically, humans always firstly determine one vehicle's coarse-grained category, i.e., the car model/type.",2433
vision Person is preliminary computer tasks. operation for several high level detection fundamental,Person detection is fundamental preliminary operation for several high level computer vision tasks.,2434
person is can One significantly retail. industry that benefit from detection,One industry that can significantly benefit from person detection is retail.,2435
"detectors efficiently in face to single-shot we videos. localize faces use multi-scale First,","First, we use multi-scale single-shot face detectors to efficiently localize faces in videos.",2436
"architecture are methods (NAS) Due search neural expensive. computationally to previous this,","Due to this, previous neural architecture search (NAS) methods are computationally expensive.",2437
factors and depends target devices. resolution architecture on as such optimality ConvNet input,ConvNet architecture optimality depends on factors such as input resolution and target devices.,2438
"for However, case-by-case redesigns. expensive too existing approaches are","However, existing approaches are too expensive for case-by-case redesigns.",2439
"In completed a pattern this propose new descriptor, paper, local we derivative texture (CLDP).","In this paper, we propose a new texture descriptor, completed local derivative pattern (CLDP).",2440
"lead of period, the behavior During may this volatility explosive an to change the process.","During this period, the change may lead to an explosive behavior of the volatility process.",2441
that assumed the the unknown. of magnitude the breaks both It are is and timing,It is assumed that both the magnitude and the timing of the breaks are unknown.,2442
moment Our hold on under innovations GARCH process. theoretical the of assumptions results mild the,Our theoretical results hold under mild moment assumptions on the innovations of the GARCH process.,2443
to Bitcoin returns returns. Apple test and the We apply asset,We apply the test to Apple asset returns and Bitcoin returns.,2444
also provide and learning-based comparison theoretical and We methods. rule-based our the greedy between experimental,We also provide the theoretical and experimental comparison between our learning-based and greedy rule-based methods.,2445
give widely look. applied artistic images has new a been real-world transfer to Style,Style transfer has been widely applied to give real-world images a new artistic look.,2446
vision component systems. Outdoor to crucial visual localization is a computer many,Outdoor visual localization is a crucial component to many computer vision systems.,2447
"relevant task image, to repository. search is query Given the in images a a","Given a query image, the task is to search relevant images in a repository.",2448
these the combining feature by Thus vectors. CBIR system be can of enhanced performance the,Thus the performance of the CBIR system can be enhanced by combining these feature vectors.,2449
"and texture color, have to four taken shape We feature represent descriptors features.","We have taken four feature descriptors to represent color, texture and shape features.",2450
peer-reviewed was network developed literature. of detailed modules from event A,A network of detailed event modules was developed from peer-reviewed literature.,2451
Compressing ever-increasing focus. (CNNs) received networks has neural research convolutional,Compressing convolutional neural networks (CNNs) has received ever-increasing research focus.,2452
"perspective. interpretable a In CNN investigate we the problem from novel paper, of compression this","In this paper, we investigate the problem of CNN compression from a novel interpretable perspective.",2453
accomplish to based KSE compression. conducted on the is Kernel CNN high-precision clustering further indicator,Kernel clustering is further conducted based on the KSE indicator to accomplish high-precision CNN compression.,2454
superior demonstrates gains over performance previous method ones. Our,Our method demonstrates superior performance gains over previous ones.,2455
widely are two SBIR on datasets. Extensive used experiments benchmark conducted,Extensive experiments are conducted on two widely used SBIR benchmark datasets.,2456
Neural modeled systems of connected neural be networks as can functionally elements.,Neural systems can be modeled as networks of functionally connected neural elements.,2457
Neural performance Convolution achieved Networks super-resolution Deep image significant (SR) have recently. (CNN) single on,Deep Convolution Neural Networks (CNN) have achieved significant performance on single image super-resolution (SR) recently.,2458
mechanisms and cellular Various the dynamics structure influence the level firing patterns. network macroscopic,Various cellular level mechanisms influence the network dynamics and structure the macroscopic firing patterns.,2459
"delay a i.e. symmetry necessary condition We transmission show that is for breaking, the","We show that the transmission delay is a necessary condition for symmetry breaking, i.e.",2460
time. space convolutions so do and deformable leveraging We across by,We do so by leveraging deformable convolutions across space and time.,2461
these Motion motion as (Discriminative refer DiMoFs We to Features). offsets,We refer to these motion offsets as DiMoFs (Discriminative Motion Features).,2462
"within grounder proposed the a given image to Subsequently, expression is directly. referring localize the","Subsequently, a grounder is proposed to localize the referring expression within the given image directly.",2463
"a and of CNNs. train collecting However, data amount annotating to huge are needed","However, collecting and annotating a huge amount of data are needed to train CNNs.",2464
"the features obtains invariant we class-wise develop which Then, of CNNs. adaptation domain","Then, we develop the class-wise adaptation which obtains domain invariant features of CNNs.",2465
approach kriging. novel based single-image a geostatistical the propose of super-resolution on method We,We propose a novel single-image super-resolution approach based on the geostatistical method of kriging.,2466
as super-resolution other results state-of-the-art methods. Our network achieves competitive,Our network achieves competitive super-resolution results as other state-of-the-art methods.,2467
interact question. oscillations How remains the salient a rate and firing,How the firing rate and oscillations interact remains a salient question.,2468
"consider not available. bounding is might inefficient, However, this solution the be we boxes since","However, we consider this solution is inefficient, since the bounding boxes might not be available.",2469
popular combines of a (e.g. object number SIMRDWN algorithms detection,SIMRDWN combines a number of popular object detection algorithms (e.g.,2470
"Super-resolution less performance. beneficial though small is at in resolutions, improvement provides a coarser still","Super-resolution is less beneficial at coarser resolutions, though still provides a small improvement in performance.",2471
face role recognition a plays Facial critical in and landmark localization analysis.,Facial landmark localization plays a critical role in face recognition and analysis.,2472
generates response facial landmark any raw without Our images preprocessing. from directly proposed BB-FCN maps,Our proposed BB-FCN generates facial landmark response maps directly from raw images without any preprocessing.,2473
"activated naturally neighboring only response long-distance function, by Gaussian is and points. penalizes","Gaussian function, naturally penalizes long-distance response and is only activated by neighboring points.",2474
localized features Such response generates discriminative point highly given distributions. different,Such localized response generates highly discriminative features given different point distributions.,2475
This for motivates a spatial-spectral denoising. HSI us propose paradigm unified to,This motivates us to propose a unified spatial-spectral paradigm for HSI denoising.,2476
annotations The of locations. is and properties accompanied spatial data by object individual training often,The training data is often accompanied by annotations of individual object properties and spatial locations.,2477
"available and obtain. expertise require scarcely to are However, and binaural nontrivial recordings equipment","However, binaural recordings are scarcely available and require nontrivial expertise and equipment to obtain.",2478
monaural convert binaural leveraging to audio common into by video. audio We propose,We propose to convert common monaural audio into binaural audio by leveraging video.,2479
from link multi-modal video. this approach Our recovers unlabeled,Our multi-modal approach recovers this link from unlabeled video.,2480
and significant incurs our real-time system for suitable makes This savings approach operation. runtime hierarchical,This hierarchical approach incurs significant runtime savings and makes our system suitable for real-time operation.,2481
toward a dermatological is identification step lesion key diagnosis. Skin,Skin lesion identification is a key step toward dermatological diagnosis.,2482
"transform In propose this we a a video to comics. a into solution paper,","In this paper, we propose a solution to transform a video into a comics.",2483
neural a based We this task using approach algorithm on Adversarial Networks style Generative (GANs).,We approach this task using a neural style algorithm based on Generative Adversarial Networks (GANs).,2484
"a into comics. the style second In selected of the is stage, keyframes transferred","In the second stage, the style of selected keyframes is transferred into a comics.",2485
broad well-researched Super Single is relevance. (SISR) Resolution  problem with  commercial a Image,  Single Image Super Resolution (SISR) is a well-researched problem with broad commercial relevance.,2486
years. Face has great the in made past progress synthesis few sketch,Face sketch synthesis has made great progress in the past few years.,2487
process. two-fold pose representation signature in done is The,The pose signature representation is done in two-fold process.,2488
images. a of set without any training recognize Zero-shot to classes (ZSL) learning unseen aims,Zero-shot learning (ZSL) aims to recognize a set of unseen classes without any training images.,2489
thus limits is annotation it of severely label/attribute ZSL. expensive; scalability the Class,Class label/attribute annotation is expensive; it thus severely limits the scalability of ZSL.,2490
ZSL is yet more conventional setting. more the clearly than This challenging realistic,This is clearly more challenging yet more realistic than the conventional ZSL setting.,2491
features followed for bidirectional This learning by is ZSL. and attributes between projections,This is followed by learning bidirectional projections between features and attributes for ZSL.,2492
"with theoretic efficient rigorous solver provided, is algorithm together An analysis.","An efficient solver is provided, together with rigorous theoretic algorithm analysis.",2493
and that experiments ZSL show our model SIA. achieves superior performance both on Extensive,Extensive experiments show that our model achieves superior performance on both ZSL and SIA.,2494
for A extracting discriminative features. Net (FDFNet) been Extraction novel Feature proposed Finger-Dorsal has the,A novel Finger-Dorsal Feature Extraction Net (FDFNet) has been proposed for extracting the discriminative features.,2495
fraction conventional fat a were maps created method. using Reference,Reference fat fraction maps were created using a conventional method.,2496
performed graphics was processing a Training unit (GPU). using,Training was performed using a graphics processing unit (GPU).,2497
unit using as (CPU). well the performed GPU a central was processing as Inference,Inference was performed using the GPU as well as a central processing unit (CPU).,2498
represents listening hearing emerging Machine or an area.,Machine hearing or listening represents an emerging area.,2499
"learning a sake, is efficient this and method For supervised novel dictionary presented.","For this sake, a novel and efficient supervised dictionary learning method is presented.",2500
"and words, the other seeking maximize separability. minimize class intra-class homogeneity In are to we","In other words, we are seeking to minimize the intra-class homogeneity and maximize class separability.",2501
resulting descent is method. solved and The optimization proximal using gradient non-convex a problem,The resulting optimization problem is non-convex and solved using a proximal gradient descent method.,2502
detection in a is surveillance. Anomaly video challenging intelligent problem,Anomaly detection is a challenging problem in intelligent video surveillance.,2503
"requirement. methods cannot which existing are Most consuming, real-time the satisfy computation","Most existing methods are computation consuming, which cannot satisfy the real-time requirement.",2504
look interesting images iris both Generating which realistic and an challenging is problem.,Generating iris images which look realistic is both an interesting and challenging problem.,2505
achieves experimental proposed Multiple the different state-of-the-art on evaluated protocols demonstrate experiments performance. that method,Multiple experiments evaluated on different experimental protocols demonstrate that the proposed method achieves state-of-the-art performance.,2506
"this data concepts, in prove and experimental the computer paper. To presented simulations are underlying","To prove the underlying concepts, computer simulations and experimental data are presented in this paper.",2507
"and and to During depth are alternated depth. motion accurate estimation converge inference,","During inference, motion and depth estimation are alternated and converge to accurate depth.",2508
"challenging in to regions, discontinuities. as such leverage This allows us depth depth uncertainty","This allows us to leverage depth uncertainty in challenging regions, such as depth discontinuities.",2509
In examine race believe cases paper one this is such where factor. we we,In this paper we examine cases where we believe race is one such factor.,2510
and image forms look facial We problem; synthesis. classification particular two in at of,We look in particular at two forms of problem; facial classification and image synthesis.,2511
of solutions to respect correspond VDA. from with to Justified the argumentation actions semantics,Justified actions with respect to semantics from argumentation correspond to solutions of the VDA.,2512
usefulness This synthetic from data information the of demonstrates semantic for segmentation. geometric cross-domain,This demonstrates the usefulness of geometric information from synthetic data for cross-domain semantic segmentation.,2513
average show in performance an Results approx. decoding increase of,Results show an average increase in decoding performance of approx.,2514
optimization They with where complex solvers problems performance. varied exhibit deal standalone,They deal with complex optimization problems where standalone solvers exhibit varied performance.,2515
selection Among tool hyper-heuristics. reside such a,Among such a tool reside selection hyper-heuristics.,2516
"the their of analyzing on transformation, each effect of combinations. focus kind of and We","We focus on analyzing the effect of each kind of transformation, and of their combinations.",2517
"on increasing previous Furthermore, analyzed features. we of research expand by number the","Furthermore, we expand on previous research by increasing the number of analyzed features.",2518
"as However, implicit explicit fruitful. transformations not and combining was","However, combining implicit and explicit transformations was not as fruitful.",2519
"some ran of we domain tests Additionally, the on problems. confirmatory knapsack","Additionally, we ran some confirmatory tests on the domain of knapsack problems.",2520
measurement. concepts Our key of on is focus the,Our focus is on the key concepts of measurement.,2521
"or the need for due system to eliminated. revisions changes settings is user-defined Hence, future","Hence, the need for user-defined settings or future revisions due to system changes is eliminated.",2522
through in simulations described scheme evaluated test The and is systems. two,The scheme is described and evaluated through simulations in two test systems.,2523
"basic each element. is It which seeds for the object point, proposal","It seeds object proposal for each point, which is the basic element.",2524
yield performance. with and context features precise coordinates improved both cloud These information point,These features with both context information and precise point cloud coordinates yield improved performance.,2525
"hard new great on set. accomplishes advantage , state-of-the-art the Our showing method","Our method accomplishes new state-of-the-art , showing great advantage on the hard set.",2526
Automated detection deception (ADD) videos from task. is a challenging real-life,Automated deception detection (ADD) from real-life videos is a challenging task.,2527
"In work, are problems addressed. this both","In this work, both problems are addressed.",2528
"novel cross-stream a is network for face-focused multimodal learning, (FFCSN) Specifically, face-body proposed.","Specifically, for face-body multimodal learning, a novel face-focused cross-stream network (FFCSN) is proposed.",2529
results. show our Extensive FFCSN model that experiments state-of-the-art achieves,Extensive experiments show that our FFCSN model achieves state-of-the-art results.,2530
"to directly learn human-object an ""hotspots"" from interaction video. We propose approach","We propose an approach to learn human-object interaction ""hotspots"" directly from video.",2531
The benefits DLOW our two-fold. model are of,The benefits of our DLOW model are two-fold.,2532
"First, intermediate it different is into in able to images the domains. styles transfer source","First, it is able to transfer source images into different styles in the intermediate domains.",2533
model implement CycleGAN. based We on DLOW our,We implement our DLOW model based on CycleGAN.,2534
"using region this network. attention we paper, challenges address guided a the In","In this paper, we address the challenges using a region guided attention network.",2535
"domain data. for unlabeled perform adaptation addition, target In we","In addition, we perform domain adaptation for unlabeled target data.",2536
and in Individual an detection tracking is monitoring requirement pig applications. many important pig video-based,Individual pig detection and tracking is an important requirement in many video-based pig monitoring applications.,2537
"this From and obtain several useful study, observations. we interesting","From this study, we obtain several interesting and useful observations.",2538
"Tree-structured Kronecker convolution employs TFA and Convolutional we Networks Finally, Kronecker design which module. (TKCN)","Finally, we design Tree-structured Kronecker Convolutional Networks (TKCN) which employs Kronecker convolution and TFA module.",2539
"and exploit attention this a important we language-guided information capture To mechanism. graph-based, propose","To capture and exploit this important information we propose a graph-based, language-guided attention mechanism.",2540
"proposed and comprehension mechanism visualisable enables attention the to the be graph decision explainable. Furthermore,","Furthermore, the proposed graph attention mechanism enables the comprehension decision to be visualisable and explainable.",2541
on show comprehension of Experiments datasets the advantage approach. referring expression the three proposed,Experiments on three referring expression comprehension datasets show the advantage of the proposed approach.,2542
"existing models the this paper, we enable same. to In video do","In this paper, we enable existing video models to do the same.",2543
This introduces and publicly recognition. in detection text article scene available datasets,This article introduces publicly available datasets in scene text detection and recognition.,2544
"on deep PReID part-based methods models. state-of-the-art Recently, are many","Recently, many state-of-the-art methods on PReID are part-based deep models.",2545
"direction feature However, usually vertical the body representation of is ignored. in","However, the feature representation of body in vertical direction is usually ignored.",2546
"model for In study, PReID. we multi-branches a introduce this deep","In this study, we introduce a multi-branches deep model for PReID.",2547
"consists Specifically, five of model branches. the","Specifically, the model consists of five branches.",2548
learning. branches the five representation improve All the can,All the five branches can improve the representation learning.,2549
proposed deep framework most cases. many outperforms state-of-the-art The in,The proposed deep framework outperforms many state-of-the-art in most cases.,2550
train initial Directly these harmful segmentation is for to model using masks performance. the,Directly using these initial masks to train segmentation model is harmful for the performance.,2551
"paper. this network hybrid problem, propose address in a To we this","To address this problem, we propose a hybrid network in this paper.",2552
has to challenge computer traditional been Scale from vision. approaches modern variation a in,Scale variation has been a challenge from traditional to modern approaches in computer vision.,2553
data. from the We should argue learned scaling that be policy,We argue that the scaling policy should be learned from data.,2554
with results scale Our challenges. improvement images show major for,Our results show major improvement for images with scale challenges.,2555
at provided the to document. code of MATLAB source end A link is the,A link to MATLAB source code is provided at the end of the document.,2556
framework the main decouples three The factors (i.e.,The framework decouples the three main factors (i.e.,2557
the could show We that segmentation improve the truth effectively pseudo ground performance.,We show that the pseudo ground truth could effectively improve the segmentation performance.,2558
than more efficient is segmentation object video methods. straightforward This existing method unsupervised,This straightforward unsupervised video object segmentation method is more efficient than existing methods.,2559
"provided are theoretical with along practical remarks. Finally, and conclusions","Finally, conclusions are provided along with theoretical and practical remarks.",2560
recognition. deals during current learning both images object in static work with recognition and Most,Most current work in object recognition deals with static images during both learning and recognition.,2561
in use We information these both the of sources recognition of and explore learning processes.,We explore the use of these sources of information in both learning and recognition processes.,2562
The can form composed the to local be together global match then density. distributions,The local distributions can then be composed together to form the global match density.,2563
makes This large particularly models. it suitable for handling,This makes it particularly suitable for handling large models.,2564
automatically more makes technology dramatically Text impressive. transfer the effects text,Text effects transfer technology automatically makes the text dramatically more impressive.,2565
recognize works important AUs regions. emphasizing attempted Current by have to,Current works have attempted to recognize AUs by emphasizing important regions.,2566
"novel model R-CNN AU knowledge, By a expert based incorporating propose named R-CNN. we prior","By incorporating expert prior knowledge, we propose a novel R-CNN based model named AU R-CNN.",2567
performance design considerably approaches. detection better existing than This produces,This design produces considerably better detection performance than existing approaches.,2568
of recognition achieved. is AU detection State-of-the-art performance,State-of-the-art recognition performance of AU detection is achieved.,2569
each corresponding aims pixel with image parsing) to its image label. annotate class semantic,image parsing) aims to annotate each image pixel with its corresponding semantic class label.,2570
improved optimization typically Segmentation Random on result labels. (MRF) Markov the initial is by Field,Segmentation result is typically improved by Markov Random Field (MRF) optimization on the initial labels.,2571
"by a framework prior designing the of lack object-oriented is information. such challenged generalized However,","However, designing such a generalized framework is challenged by the lack of object-oriented prior information.",2572
almost enables improve us This with loss. no performance method accuracy to,This enables us to improve method accuracy with almost no performance loss.,2573
one issues of instantiation in to these for ontology address We TDD aim authoring.,We aim to address these issues in one instantiation of TDD for ontology authoring.,2574
"video problem action recognition, Video understanding, attention. gaining in increasing a critical been has","Video action recognition, a critical problem in video understanding, has been gaining increasing attention.",2575
"on performance and show results ActivityNet, benchmark Experimental datasets, competitive our Something-Something of method. the","Experimental results on benchmark datasets, Something-Something and ActivityNet, show the competitive performance of our method.",2576
Recent distributions. times have systems often that have fixation in such studies right-skewed found,Recent studies have found that fixation times in such systems often have right-skewed distributions.,2577
"depend their on how mutant about and fitness. however, Little known, these distributions is skew","Little is known, however, about how these distributions and their skew depend on mutant fitness.",2578
"Also, learning for architectures quantification. proposed deep were new three LV","Also, three new deep learning architectures were proposed for LV quantification.",2579
"may approach applications this quantification. important have Also, motion on","Also, this approach may have important applications on motion quantification.",2580
"more that incumbents. resources, architecture is spatio-spectral our misdetection of with It lower identifies shown","It is shown that our architecture identifies more spatio-spectral resources, with lower misdetection of incumbents.",2581
"IoT incumbents. with a As devices into are interference limited connected result, more","As a result, more IoT devices are connected with limited interference into incumbents.",2582
performing for novel proposals a propose generating We approach for region face-detection.,We propose a novel approach for generating region proposals for performing face-detection.,2583
of signals reside on applications In graphs. vertices many the weighted,In many applications signals reside on the vertices of weighted graphs.,2584
dimensionality graph learning. have focused reduction unsupervised methods dictionary on Existing signals for,Existing unsupervised dimensionality reduction methods for graph signals have focused on dictionary learning.,2585
"Action the boxes. as Machine, inputs person cropped videos The method, by takes bounding called","The method, called Action Machine, takes as inputs the videos cropped by person bounding boxes.",2586
flexibility. competitive and Results performance improved show,Results show competitive performance and improved flexibility.,2587
and characteristic regulation is processes intracellular such gene signalling. of as chemical key Stochasticity a,Stochasticity is a key characteristic of intracellular processes such as gene regulation and chemical signalling.,2588
phenomena. of capture be reacting to able stochastic must biochemically Mathematical idealisations systems,Mathematical idealisations of biochemically reacting systems must be able to capture stochastic phenomena.,2589
decades. been research These challenges for of active over four have computational subjects,These computational challenges have been subjects of active research for over four decades.,2590
implementations. are MATLAB methods with Detailed and algorithms particularly for complemented described important,Detailed algorithms for particularly important methods are described and complemented with MATLAB implementations.,2591
a Machine transferred extracted classification. Features and to Support are for Vector Linear,Features are extracted and transferred to a Linear Support Vector Machine for classification.,2592
available and datasets are as performed All JAFFE such CK+ publicly two database. on experiments,All experiments are performed on two publicly available datasets such as JAFFE and CK+ database.,2593
"saliency information pixel saliency In make contextual to detection, needs every prediction.","In saliency detection, every pixel needs contextual information to make saliency prediction.",2594
We PiCANets the salient proposed for architecture into object a detection. U-Net introduce,We introduce the proposed PiCANets into a U-Net architecture for salient object detection.,2595
performance. PiCANets Experimental saliency improve the significantly that can results detection the proposed indicate,Experimental results indicate that the proposed PiCANets can significantly improve the saliency detection performance.,2596
"Consequently, our model performs favorably other saliency against methods. state-of-the-art","Consequently, our saliency model performs favorably against other state-of-the-art methods.",2597
contributions. three upon Our builds main approach,Our approach builds upon three main contributions.,2598
updates and Our set model re-ID iteratively. the learns annotation method the,Our method learns the re-ID model and updates the annotation set iteratively.,2599
"simple, proposed three method being the effectiveness its demonstrates datasets. person While re-ID on video-based","While being simple, the proposed method demonstrates its effectiveness on three video-based person re-ID datasets.",2600
the core of Minimally linkages frustrated highlight folding molecule. stable the,Minimally frustrated linkages highlight the stable folding core of the molecule.,2601
artists Deep to and tool becoming generation designers creativity a image is enhance potential.,Deep image generation is becoming a tool to enhance artists and designers creativity potential.,2602
advantages presents convolutional This of used compared the number a to commonly architectures. network,This presents a number of advantages compared to the commonly used convolutional network architectures.,2603
"(depth, etc.) flow, learning geometric for optical Unsupervised perception","Unsupervised learning for geometric perception (depth, optical flow, etc.)",2604
to interest great autonomous is systems. of,is of great interest to autonomous systems.,2605
"are robust, which learning large In dataset. supervised contrast, labeled geometric require algorithms,","In contrast, supervised learning algorithms, which are robust, require large labeled geometric dataset.",2606
atrial of fibrillation. volume left MR ablation benefit the procedure can segmenting Accurately atrium in,Accurately segmenting left atrium in MR volume can benefit the ablation procedure of atrial fibrillation.,2607
labor-intensive often experts automated in Traditional manual relieving labeling. solutions the from fail,Traditional automated solutions often fail in relieving experts from the labor-intensive manual labeling.,2608
"a of training competitive the to scheme we further Finally, ability networks. generalization boost devise","Finally, we devise a competitive training scheme to further boost the generalization ability of networks.",2609
to demonstrate We model's also ability the our rank-aware of video. parts to attend,We also demonstrate our model's ability to attend to rank-aware parts of the video.,2610
"complicated design hardware the for is implementation. and joint inappropriate However,","However, the joint design is complicated and inappropriate for hardware implementation.",2611
"implementation-friendly orthogonal structure. optimal modulation, LMMSE For has an receiver the","For orthogonal modulation, the optimal LMMSE receiver has an implementation-friendly structure.",2612
left of localized atrium firstly automatically by region detection The module. a is,The region of left atrium is firstly automatically localized by a detection module.,2613
with scheme. obtained Further improvement recursive is training an,Further improvement is obtained with an recursive training scheme.,2614
"introduced With to all studies, be effective. the ablation prove modules","With ablation studies, all the introduced modules prove to be effective.",2615
"not since feasible is can in size. be petabytes datasets analysis However, manual","However, manual analysis is not feasible since datasets can be petabytes in size.",2616
"process samples. training sampling, Such resulting stimulates balanced negative more in hard","Such process stimulates hard negative sampling, resulting in more balanced training samples.",2617
"the more distinguishing background Consequently, discriminative RPNs (i.e., distractors). are sequentially in difficult similar","Consequently, the RPNs are sequentially more discriminative in distinguishing difficult background (i.e., similar distractors).",2618
function. the is with trained multi-task end-to-end loss C-RPN,C-RPN is trained end-to-end with the multi-task loss function.,2619
"adaption, is inference, tracking. In it without C-RPN as real-time deployed any for is, temporal","In inference, C-RPN is deployed as it is, without any temporal adaption, for real-time tracking.",2620
"of results various have learning fields approaches in vision. computer promising Recently, achieved deep","Recently, deep learning approaches have achieved promising results in various fields of computer vision.",2621
"challenging is applications Nowadays, task VCO high for a different engineers. for performance a designing","Nowadays, designing a high performance VCO for different applications is a challenging task for engineers.",2622
of friends everyday Millions perform and impaired to visually depend tasks. on relatives people their,Millions of visually impaired people depend on relatives and friends to perform their everyday tasks.,2623
and is available Pay free. Play for App publicly Voice on Store Google,Pay Voice is publicly available on Google Play and App Store for free.,2624
Social relationships of humans. basis social the structure of form,Social relationships form the basis of social structure of humans.,2625
"the in problem recognition study images. we social this work, of In visual relationship","In this work, we study the problem of visual social relationship recognition in images.",2626
frequencies. in One obstacle Theory to its applying relationship is Dempster-Shafer important (DST),One important obstacle in applying Dempster-Shafer Theory (DST) is its relationship to frequencies.,2627
"particular, belief difficulties in functions factorizations of from In serious data. finding exist there","In particular, there exist serious difficulties in finding factorizations of belief functions from data.",2628
difficulty. attempt a is this overcome to In made novel this paper,In this paper a novel attempt is made to overcome this difficulty.,2629
environment. due a to appearance and unconstrained dynamic is variations problem complex Tracking Visual,Visual Tracking is a complex problem due to unconstrained appearance variations and dynamic environment.,2630
accidents remain of and issues driving. important of road traffic autonomous decline Safety,Safety and decline of road traffic accidents remain important issues of autonomous driving.,2631
of evaluated. segmentation model and We are usability is presented using the is color-based the,We are using color-based segmentation is presented and the usability of the model is evaluated.,2632
to A refinement residual-based is Adam performance. state-of-art boundary achieve used optimization and also,A residual-based boundary refinement and Adam optimization is also used to achieve state-of-art performance.,2633
it in world. get real We propose framework work a to,We propose a framework to get it work in real world.,2634
because photography enjoy appreciate food. People food they,People enjoy food photography because they appreciate food.,2635
models available. and We publicly make code,We make code and models publicly available.,2636
(ZSL) its to Zero-shot learning see the without a solving need enables examples. task,Zero-shot learning (ZSL) enables solving a task without the need to see its examples.,2637
mutual can without supervision. inter-connected synthesizers trained are two end-to-end as The be which references,The two synthesizers are inter-connected as mutual references which can be trained end-to-end without supervision.,2638
SF-GAN. Qualitative demonstrate superiority state-of-the-art the of proposed quantitative the and the comparisons with,Qualitative and quantitative comparisons with the state-of-the-art demonstrate the superiority of the proposed SF-GAN.,2639
only the overcomes negative bags. using It the limitations voting of,It overcomes the limitations of voting using only the negative bags.,2640
beyond Experimental superiority demonstrate our the scheme baselines. of results the,Experimental results demonstrate the superiority of our scheme beyond the baselines.,2641
of The are three-fold. this contributions work,The contributions of this work are three-fold.,2642
"of performance consolidated If help approaches? current multiple samples actions can from improve so, the","If so, can consolidated samples from multiple actions help improve the performance of current approaches?",2643
brain in often test used test an is The procedure permutation imaging.,The permutation test is an often used test procedure in brain imaging.,2644
effectiveness results the demonstrated method. of Experimental have our,Experimental results have demonstrated the effectiveness of our method.,2645
in with motor control disease enhance Parkinson's shown to has Music (PD). patients been,Music has been shown to enhance motor control in patients with Parkinson's disease (PD).,2646
perception and movement its music improve PD? associating Would to in control,Would associating music to movement improve its perception and control in PD?,2647
The validation in PD of patients. method underway is handwriting for a such,The validation of such a method is underway for handwriting in PD patients.,2648
rely also milder approach on assumptions that difference-in-discontinuities are the fuzzy Modifications of proposed.,Modifications of the fuzzy difference-in-discontinuities approach that rely on milder assumptions are also proposed.,2649
be machines on question the whether This paper rational. studies can,This paper studies the question on whether machines can be rational.,2650
of machines. humans are transferred It to the of these studies limitations whether limitations,It studies whether these limitations of humans are transferred to the limitations of machines.,2651
humans. It that than also machines more can concludes be rational,It also concludes that machines can be more rational than humans.,2652
use The saturation algorithm for personal algorithm temperature (NIPST) comparisons. was partly,The partly personal saturation temperature (NIPST) algorithm was use for algorithm comparisons.,2653
(DL) also generated SSI deep algorithm was learning for without Another algorithm comparisons.,Another deep learning algorithm without SSI (DL) was also generated for algorithm comparisons.,2654
superiority and effectiveness approach. of our the results Extensive validate,Extensive results validate the effectiveness and superiority of our approach.,2655
be public upon acceptance. will paper Code made,Code will be made public upon paper acceptance.,2656
quantifies in the model. a of sensitivity variables uncertain influence analysis mathematical Global (GSA),Global sensitivity analysis (GSA) quantifies the influence of uncertain variables in a mathematical model.,2657
be typically knowledge. This distribution must chosen limited and is unknown using and/or data,This distribution is typically unknown and must be chosen using limited data and/or knowledge.,2658
of uncertainty. indices their robustness Sobol' to the depends The this usefulness distributional on,The usefulness of the Sobol' indices depends on their robustness to this distributional uncertainty.,2659
contaminant transport. illustrated through The for and a method synthetic is examples model,The method is illustrated through synthetic examples and a model for contaminant transport.,2660
optimization a layer-by-layer Most in obtained mode. results of are state-of-the-art,Most of state-of-the-art results are obtained in a layer-by-layer optimization mode.,2661
our and evaluations superior accuracy of comparisons show The and method. effectiveness,The evaluations and comparisons show superior effectiveness and accuracy of our method.,2662
"of edges lead blurred plausible due to losses averaging to minimizing Unfortunately, solutions. these inevitably","Unfortunately, minimizing these losses inevitably lead to blurred edges due to averaging of plausible solutions.",2663
"sharp our Qualitatively, other proposed methods. compared with state-of-the-art method yielded edges to SR images","Qualitatively, our proposed method yielded images with sharp edges compared to other state-of-the-art SR methods.",2664
Scene has research text increased community. the in attention recognition received,Scene text recognition has received increased attention in the research community.,2665
progressively to text performance. recognition calibrates The boost RCN the irregular the,The RCN progressively calibrates the irregular text to boost the recognition performance.,2666
"way, preserved original transformation. the this character information final In the is until","In this way, the original character information is preserved until the final transformation.",2667
designs succeeding the results Such performance recognition. lead optimal to of to boost calibration,Such designs lead to optimal calibration results to boost the performance of succeeding recognition.,2668
"superiority our method, the on challenging benchmarks. demonstrate Extensive irregular datasets on experiments especially of","Extensive experiments on challenging datasets demonstrate the superiority of our method, especially on irregular benchmarks.",2669
"Chinese (TCM), diagnosis In basis are facial Traditional features important Medicine treatment. and for","In Traditional Chinese Medicine (TCM), facial features are important basis for diagnosis and treatment.",2670
works and to analyze according prescription generate symptoms. Previous,Previous works analyze and generate prescription according to symptoms.,2671
challenging is job. a novel and It,It is a novel and challenging job.,2672
much Comparison state-of-the-art. PointNetVLAD LPD-Net results reaches is better show the that and than our,Comparison results show that our LPD-Net is much better than PointNetVLAD and reaches the state-of-the-art.,2673
"occlusion, blur, to variations address synthetic customized We (scale, types specific datasets etc. pose, of","We customized synthetic datasets to address specific types of variations (scale, pose, occlusion, blur, etc.",2674
"performances. systematically investigate detection influence the different on and ), of face variations","), and systematically investigate the influence of different variations on face detection performances.",2675
detection these contribute to better how and whether factors examine face performances. We,We examine whether and how these factors contribute to better face detection performances.,2676
Edge-preserving computer is applications. vision fundamental for procedure many graphic a and smoothing,Edge-preserving smoothing is a fundamental procedure for many computer vision and graphic applications.,2677
This achieved methods. with be methods either or global can local,This can be achieved with either local methods or global methods.,2678
"global performance over most can methods superior cases, ones. In yield local","In most cases, global methods can yield superior performance over local ones.",2679
"ones. global than However, run local much usually faster methods","However, local methods usually run much faster than global ones.",2680
can proposed method The global method. show with performance comparable the state-of-the-art,The proposed method can show comparable performance with the state-of-the-art global method.,2681
to can further more They applications. be modified handle,They can be further modified to handle more applications.,2682
"extracted expression-related features Then, from nodes. leaf are","Then, expression-related features are extracted from leaf nodes.",2683
networks successful variety prediction been have a of tasks. Deep neural (DNNs) enormously across,Deep neural networks (DNNs) have been enormously successful across a variety of prediction tasks.,2684
in real-time security-sensitive systems. This widely makes deployable Defense-VAE,This makes Defense-VAE widely deployable in real-time security-sensitive systems.,2685
proposed two the of the on datasets. RGB-D We publicly effectiveness demonstrate available network,We demonstrate the effectiveness of the proposed network on two publicly available RGB-D datasets.,2686
"not well cross-modality been during between However, coherence and audio synthesis. information addressed video has","However, cross-modality coherence between audio and video information has not been well addressed during synthesis.",2687
perception describing phenomena surrounding neuro-physiological We contrast induced a by differential orientations. model consider,We consider a differential model describing neuro-physiological contrast perception phenomena induced by surrounding orientations.,2688
different detection we algorithms. representation this With LiDAR-based can existing apply,With this representation we can apply different existing LiDAR-based detection algorithms.,2689
"paper present scenario, different In we model a this building.","In this paper we present a different scenario, model building.",2690
"model contrast underlying i.e. the to In algorithm, usage","In contrast to model usage the underlying algorithm, i.e.",2691
"parameters associated also changes therefore and the underlying model, change. the","the underlying model, changes and therefore the associated parameters also change.",2692
to features. quantify objective is important these it Therefore that measures find,Therefore it is important to find objective measures that quantify these features.,2693
analysis parameter input output sampling the by an space. Our based on set generated is,Our analysis is based on an output set generated by sampling the input parameter space.,2694
an present Hence techniques analyze ensemble metrics we probabilistic also of volumes. and to,Hence we also present techniques and metrics to analyze an ensemble of probabilistic volumes.,2695
guidance. the rely the to on provide user as image annotations algorithms Interactive segmentation,Interactive image segmentation algorithms rely on the user to provide annotations as the guidance.,2696
sparsity the of query diversity selection on We the criteria enforce the and seeds.,We enforce the sparsity and diversity criteria on the selection of the query seeds.,2697
algorithm our publicly performance available six The of evaluated datasets. on is,The performance of our algorithm is evaluated on six publicly available datasets.,2698
"trajectories representation form. introduce a high level we gradient object using of First, color","First, we introduce a high level representation of object trajectories using color gradient form.",2699
Stochastic t-Distributed Neighbor are Anomalous trajectories (t-SNE). Embedding separated using,Anomalous trajectories are separated using t-Distributed Stochastic Neighbor Embedding (t-SNE).,2700
"has for been a hybrid trajectory and detection. architecture used CNN-VAE anomaly classification Finally,","Finally, a hybrid CNN-VAE architecture has been used for trajectory classification and anomaly detection.",2701
TX the distances were horizontal RXs Different the and between used measurements. in vertical and,Different horizontal and vertical distances between the RXs and TX were used in the measurements.,2702
input. faithful the to the It makes reconstruction,It makes the reconstruction faithful to the input.,2703
helps loss to multiple from solutions. most plausible appearance select Adversarial the,Adversarial loss helps to select the most plausible appearance from multiple solutions.,2704
a incentives. positive and of mix negative work experiment entails This,This experiment entails a mix of positive and negative work incentives.,2705
for supply evidence provide report I methods theoretical can CML the that predictions. support labor,I report evidence that CML methods can provide support for the theoretical labor supply predictions.,2706
the is Mixture forthcoming This Analysis'. a 'Handbook for of prepared chapter,This is a chapter prepared for the forthcoming 'Handbook of Mixture Analysis'.,2707
ability Continual improve of capacities to throughout is continually. the tasks their learning agents multiple,Continual learning is the ability of agents to improve their capacities throughout multiple tasks continually.,2708
mentioning. agent worth features are our Two critical of,Two critical features of our agent are worth mentioning.,2709
"not from our any Firstly, agent did learn demonstrations.","Firstly, our agent did not learn from any demonstrations.",2710
in paper systems. receive multiple-input (MIMO) massive antenna selection multiple-output studies This,This paper studies receive antenna selection in massive multiple-input multiple-output (MIMO) systems.,2711
"the nuclei Additionally, clustered together. of are some often","Additionally, some of the nuclei are often clustered together.",2712
to on demonstrate methods We independent multi-tissue state-of-the-art histology multiple image performance compared datasets. other,We demonstrate state-of-the-art performance compared to other methods on multiple independent multi-tissue histology image datasets.,2713
has area. In recent tracking research years object very become active visual a,In recent years visual object tracking has become a very active research area.,2714
being of proposed each algorithms An increasing tracking are number year.,An increasing number of tracking algorithms are being proposed each year.,2715
comprises first The survey trackers. the comprehensive recently proposed work of a part of this,The first part of this work comprises a comprehensive survey of the recently proposed trackers.,2716
Non-CFTs. Trackers We and broadly categorize Correlation trackers based into (CFTs) Filter,We broadly categorize trackers into Correlation Filter based Trackers (CFTs) and Non-CFTs.,2717
Our others. the concludes Discriminative based than better Filter Correlation trackers that (DCF) perform study,Our study concludes that Discriminative Correlation Filter (DCF) based trackers perform better than the others.,2718
Two been proposed explain of types have mechanisms this to conundrum.,Two types of mechanisms have been proposed to explain this conundrum.,2719
"currents. Unexpectedly, the timescale network in not adaptive do inherit dynamics large macroscopic present the","Unexpectedly, the macroscopic network dynamics do not inherit the large timescale present in adaptive currents.",2720
is semantic auxiliary for task as navigation. applied segmentation Image an,Image semantic segmentation is applied as an auxiliary task for navigation.,2721
by an to module optical introducing information network. consider We and LSTM flow temporal the,We consider temporal information by introducing an LSTM module and optical flow to the network.,2722
"with combine kinematics we Finally, a vehicle sensor step. fusion","Finally, we combine vehicle kinematics with a sensor fusion step.",2723
segmentation in Errors propagate stages. to later,Errors in segmentation propagate to later stages.,2724
to segmentation are approaches specific Current tuned environments.,Current segmentation approaches are tuned to specific environments.,2725
using convolution We iris propose for neural a segmentation. network,We propose using a convolution neural network for iris segmentation.,2726
accurate algorithm multiple when a is environments. environment single tested trained Our and in in,Our algorithm is accurate when trained in a single environment and tested in multiple environments.,2727
Our the network. previous R-CNN including approaches segments faster Mask than approach,Our approach segments faster than previous approaches including the Mask R-CNN network.,2728
Alzheimer's remains of disease relation be Their to to studied. markers pathophysiological,Their relation to pathophysiological markers of Alzheimer's disease remains to be studied.,2729
"applications we in Then, widespread remote sensing. provide an of the overview UAVs of","Then, we provide an overview of the widespread applications of UAVs in remote sensing.",2730
"work for Finally, prospects are discussed. some future","Finally, some prospects for future work are discussed.",2731
awareness systems driving is A changing situational environments. enabling in dynamically for critical requirement automated,A critical requirement for automated driving systems is enabling situational awareness in dynamically changing environments.,2732
solution A possible infrastructure. by is adopt collaborative assisted possibly amongst vehicles sensing to,A possible solution is to adopt collaborative sensing amongst vehicles possibly assisted by infrastructure.,2733
analysis new performance for This networking. vehicular collaborative introduces paper and and models sensing,This paper introduces new models and performance analysis for vehicular collaborative sensing and networking.,2734
phase noise in (NLPN) systems. nonlinear multi-subcarrier correlated We explore,We explore correlated nonlinear phase noise (NLPN) in multi-subcarrier systems.,2735
Our of texture task the investigation using feasibility proves accomplishing attributes. this,Our investigation proves the feasibility of accomplishing this task using texture attributes.,2736
"associated disadvantages advantages and study, also attribute. Through the each we with identify","Through the study, we also identify advantages and disadvantages associated with each attribute.",2737
"an performance GenHap of HapCol, We compared state-of-the-art algorithm haplotype the for against phasing. efficient","We compared the performance of GenHap against HapCol, an efficient state-of-the-art algorithm for haplotype phasing.",2738
"we real assessed performance two datasets. the GenHap on Finally, of different","Finally, we assessed the performance of GenHap on two different real datasets.",2739
"color each independently. Traditionally, channel image performed interpolation warping is in color","Traditionally, color image warping interpolation is performed in each color channel independently.",2740
method our efficiency The validated extensive effectiveness of by are experiments. and,The effectiveness and efficiency of our method are validated by extensive experiments.,2741
similar to maps The latent deformations an close encoding each other variable in space. model,The latent variable model maps similar deformations close to each other in an encoding space.,2742
inference. variational on based Our unsupervised method is,Our unsupervised method is based on variational inference.,2743
a also as that regularization We filters. diffusion-based spatial such formulation present includes,We also present a formulation that includes spatial regularization such as diffusion-based filters.,2744
"velocity Additionally, estimations. multi-scale field provides framework our","Additionally, our framework provides multi-scale velocity field estimations.",2745
algorithms gains medical performance deep classification enabled significant have in learning Supervised tasks. image,Supervised deep learning algorithms have enabled significant performance gains in medical image classification tasks.,2746
that on annotation. expert these methods rely require labeled large resource-intensive datasets But,But these methods rely on large labeled datasets that require resource-intensive expert annotation.,2747
"implicitly method Further, interpretation the SSL enables predictions. of our","Further, our method implicitly enables interpretation of the SSL predictions.",2748
generative (GANs). a We for paradigm present training learning stage-wise new adversarial networks,We present a new stage-wise learning paradigm for training generative adversarial networks (GANs).,2749
this call method proposed the RankGAN. We,We call this proposed method the RankGAN.,2750
propose discriminator. GAN We first for margin-based loss the a,We first propose a margin-based loss for the GAN discriminator.,2751
often are point process the in Imperfection noise. clouds acquisition corrupted means that with,Imperfection in the acquisition process means that point clouds are often corrupted with noise.,2752
blue alternately. optimize coordinates red the then We and nodes',We then optimize the red and blue nodes' coordinates alternately.,2753
methods local performance our proposed among denoising algorithms. state-of-the-art show using experiments Extensive,Extensive experiments show state-of-the-art denoising performance among local methods using our proposed algorithms.,2754
"remove method present paper, rain we In efficient to an this streaks.","In this paper, we present an efficient method to remove rain streaks.",2755
hence representation. Cortical measures spherical in modeled be neuroimaging can,Cortical neuroimaging measures hence can be modeled in spherical representation.,2756
Free and is frames assumed. pattern of uncontrolled) (unrestricted between motion,Free (unrestricted and uncontrolled) pattern of motion between frames is assumed.,2757
"In made been deep recent strides tremendous learning. in thanks detection face have to year,","In recent year, tremendous strides have been made in face detection thanks to deep learning.",2758
"faces the as smaller. become face dramatically published detectors However, deteriorate most","However, most published face detectors deteriorate dramatically as the faces become smaller.",2759
"to Third, in multi-scale detection strategy performance enhance face and practice. testing training we use","Third, we use multi-scale training and testing strategy to enhance face detection performance in practice.",2760
"detection performance, show significantly that SFA experiments especially faces. face on small improves Comprehensive","Comprehensive experiments show that SFA significantly improves face detection performance, especially on small faces.",2761
available models will code community. Both research our and be to the,Both our code and models will be available to the research community.,2762
"The developed was (Gorban basic purposeful recently of Ecol. and kinesis Cabukoglu, model","The basic model of purposeful kinesis was developed recently (Gorban and Cabukoglu, Ecol.",2763
increasing. coefficient model the drops the diffusion this reproduction is According while to,According to this model the diffusion drops while the reproduction coefficient is increasing.,2764
"impact model of kinesis waves. purposeful this running In study, the we analyse the on","In this study, we analyse the impact of the purposeful kinesis model on running waves.",2765
with waves kinesis running experiments. in studied using numerical the population are The,The running waves in the population with kinesis are studied using numerical experiments.,2766
"prey the is the the gets population decreasing, wave While amplitude smaller. of kinesis","While the kinesis of the prey population is decreasing, the wave amplitude gets smaller.",2767
"On sharp the other for kinesis increase. lower hand, predators of have the they a","On the other hand, for the lower kinesis of predators they have a sharp increase.",2768
"refined work, simultaneously body were training and PAFs stages. previous location across In part estimation","In previous work, PAFs and body part location estimation were refined simultaneously across training stages.",2769
clip. in video when a given and objects tremendously interact challenging It becomes multiple occur,It becomes tremendously challenging when multiple objects occur and interact in a given video clip.,2770
dependencies video spatial-temporal DOA capturing in sequences consistency from frames. background adjacent models by,DOA models spatial-temporal consistency in video sequences by capturing background dependencies from adjacent frames.,2771
state-of-the-art on convolutional the various demonstrated networks Deep computing image tasks. medical have performance,Deep convolutional networks have demonstrated the state-of-the-art performance on various medical image computing tasks.,2772
the for modalities same analysis holds images from Leveraging task clinical benefits. different,Leveraging images from different modalities for the same analysis task holds clinical benefits.,2773
The with excellent experimental demonstrate efficacy the studies PnP-AdaNet. proposed comprehensive ablation of our results,The experimental results with comprehensive ablation studies demonstrate the excellent efficacy of our proposed PnP-AdaNet.,2774
upper and adjacent We with distance limit elements between lower the every thresholds. two predefined,We limit the distance between every two adjacent elements with predefined upper and lower thresholds.,2775
"in can these kdPCA recover still components examples. interpretable However,","However, kdPCA can still recover interpretable components in these examples.",2776
"populations I during Finally, using of examples two decision-making kdPCA perceptual tasks. demonstrate neural recorded","Finally, I demonstrate kdPCA using two examples of neural populations recorded during perceptual decision-making tasks.",2777
problem optimum in considers of paper the reconstruction (GSRPs). This generalized processes sampling-reconstruction,This paper considers the problem of optimum reconstruction in generalized sampling-reconstruction processes (GSRPs).,2778
one of Cross-database and emerging micro-expression micro-expression interesting analysis. in is problem (CDMER) recently recognition,Cross-database micro-expression recognition (CDMER) is one of recently emerging and interesting problem in micro-expression analysis.,2779
"paper, contribute we this In aspects. from three to this topic","In this paper, we contribute to this topic from three aspects.",2780
regression classification forest is and random the these techniques One approach. of,One of these classification and regression techniques is the random forest approach.,2781
scalability. are best for predictors tree Those decision good their computational known based performance and,Those decision tree based predictors are best known for their good computational performance and scalability.,2782
"proposed. a In work, this WEKA random is balanced forest approach for","In this work, a balanced random forest approach for WEKA is proposed.",2783
cumbersome process. is and inefficient This a,This is a cumbersome and inefficient process.,2784
"needs be once FNN to proposed trained only method. the paper, this In with","In this paper, FNN only needs to be trained once with the proposed method.",2785
quality the of the determines overall performance. rain detection An accurate streaks of,An accurate detection of rain streaks determines the quality of the overall performance.,2786
method effectiveness We quantitative and validate qualitative of evaluations. through the our,We validate the effectiveness of our method through quantitative and qualitative evaluations.,2787
our make dataset will code publicly and available. We,We will make our dataset and code publicly available.,2788
dictionary Standard due MRF to lacks computational reconstructs using maps parametric inefficiency. matching scalability and,Standard MRF reconstructs parametric maps using dictionary matching and lacks scalability due to computational inefficiency.,2789
"dictionary-matching based reduced addition, approaches. time significantly In reconstruction compared is to the","In addition, the reconstruction time is significantly reduced compared to dictionary-matching based approaches.",2790
underneath is in base common void defect roads. semi-rigid a The,The void underneath semi-rigid base is a common defect in roads.,2791
blur motion remove in event is attempt to motion to any used image. the This,This motion is used to attempt to remove any motion blur in the event image.,2792
inference inverted order in processes. models gain against to are modeled on observations Mechanistic simulation,Mechanistic simulation models are inverted against observations in order to gain inference on modeled processes.,2793
"discrepancies, model This i.e. in mismatch results","This mismatch results in model discrepancies, i.e.",2794
and predictions. systematic observations differences between model,systematic differences between observations and model predictions.,2795
"When properly, accounted discrepancies for uncertainty are is posterior not underestimated.","When discrepancies are not accounted for properly, posterior uncertainty is underestimated.",2796
need the impedes This be of process identification to formulations weak improved. that,This impedes the identification of weak process formulations that need to be improved.,2797
identified approach discrepancy GP in streams. correctly model The data rich,The GP approach correctly identified model discrepancy in rich data streams.,2798
successful to several sampling data Innovations records. application of thousand allowed observation in streams,Innovations in sampling allowed successful application to observation data streams of several thousand records.,2799
"formulation be proposed Moreover, optimization. could the new with gradient-based combined","Moreover, the proposed new formulation could be combined with gradient-based optimization.",2800
"address data two develop we augmentation techniques. issue, this To","To address this issue, we develop two data augmentation techniques.",2801
The other drone methodology create augmentation images. adversarial data to is exploiting thermal an,The other is exploiting an adversarial data augmentation methodology to create thermal drone images.,2802
"the To between we image drone, track residual a information small frames. consecutive flying utilize","To track a small flying drone, we utilize the residual information between consecutive image frames.",2803
to developed clustering agglomerative then Our hierarchical comparing validated clustering. mathematical is algorithm by morphology-based,Our developed mathematical morphology-based clustering algorithm is then validated by comparing to agglomerative hierarchical clustering.,2804
"inverters some capacities practical applications. different may power available source be cascaded in of However,","However, available source power capacities of cascaded inverters may be different in some practical applications.",2805
"autonomous consistency achieve control power is phase to proposed a voltage Moreover, synchronization. factor","Moreover, a power factor consistency control is proposed to achieve autonomous voltage phase synchronization.",2806
cascaded illustrated. are analysis and The of inverters synchronization mechanism steady-state,The steady-state analysis and synchronization mechanism of cascaded inverters are illustrated.,2807
control is The effectiveness the tested by proposed experiments. of,The effectiveness of the proposed control is tested by experiments.,2808
images. same the two and identity) mismatched,the same identity) and two mismatched images.,2809
all that have (CSPs) Problems typically many Satisfaction solutions Constraint satisfy constraints.,Constraint Satisfaction Problems (CSPs) typically have many solutions that satisfy all constraints.,2810
"some are though, solutions is, that solutions. Often preferred over some other dominate solutions others,","Often though, some solutions are preferred over others, that is, some solutions dominate other solutions.",2811
solution We reason formal as such a settings. framework present dominance about to,We present solution dominance as a formal framework to reason about such settings.,2812
"Inter-organ but spatial, only and are not functional also relations physiological.","Inter-organ relations are not only spatial, but also functional and physiological.",2813
first the k- constrained error evaluate of without We means technique coding. performance clustering,We first evaluate the error performance of k- means clustering technique without constrained coding.,2814
fractures radius common fractures are most upper of Distal in the the humans. extremity,Distal radius fractures are the most common fractures of the upper extremity in humans.,2815
is the radiologists. more detection This significantly physicians accurate and achieved by than,This is significantly more accurate than the detection achieved by physicians and radiologists.,2816
"singly the have domain been to shifting spatial investigated. to networks), or complexity connected","to singly connected networks), or shifting the complexity to spatial domain have been investigated.",2817
"limited the is optics by physically characteristics. and resolution sensor However,","However, resolution is physically limited by the optics and sensor characteristics.",2818
"of In the addition, or practical use hardware. and out-dated reasons often low-cost stipulate economic","In addition, practical and economic reasons often stipulate the use of out-dated or low-cost hardware.",2819
frames to algorithms high-resolution Multi-frame multiple images. task this fusing reconstruct by approach low-resolution,Multi-frame algorithms approach this task by fusing multiple low-resolution frames to reconstruct high-resolution images.,2820
along new This medical imaging. with work novel applications covers in super-resolution methods,This work covers novel super-resolution methods along with new applications in medical imaging.,2821
our with Practical provided. theory compatible asymptotic are also choices penalty,Practical penalty choices compatible with our asymptotic theory are also provided.,2822
requiring proposed for inference allow valid The property. without estimator oracle,The proposed estimator allow for valid inference without requiring oracle property.,2823
use image prior content and to It cameras does not synchronizes capture.,It does not use image content and synchronizes cameras prior to capture.,2824
Analytics technologies be and categorised broadly could Artificial into (AI) Autonomy. Intelligence,Artificial Intelligence (AI) technologies could be broadly categorised into Analytics and Autonomy.,2825
"Analytics knowledge algorithms of projection from gleaned on offering and perception, focuses comprehension, sensorial data.","Analytics focuses on algorithms offering perception, comprehension, and projection of knowledge gleaned from sensorial data.",2826
"around influencing shaping through the and action production. and environment revolves making, Autonomy decision","Autonomy revolves around decision making, and influencing and shaping the environment through action production.",2827
"To that requires be useful, must be trusted SAS and testing.","To be useful, SAS must be trusted and that requires testing.",2828
a SAS Lifelong learning the testing process. compounds of,Lifelong learning of a SAS compounds the testing process.,2829
"of the The to extends cons and pros, proposed concept. the challenges, discussion","The discussion extends to the challenges, pros, and cons of the proposed concept.",2830
pattern popular Deep nowadays vision recognition computer becoming in many networks tasks. are and,Deep networks are nowadays becoming popular in many computer vision and pattern recognition tasks.,2831
"of the in kernel computation issue we paper, deep networks. this address efficient In","In this paper, we address the issue of efficient computation in deep kernel networks.",2832
these kernels. deep that a dramatically complexity the novel evaluating framework reduces of propose We,We propose a novel framework that reduces dramatically the complexity of evaluating these deep kernels.,2833
paper image presents boundary. This for technique smoothing digital iterative an of approximation polygonal,This paper presents an iterative smoothing technique for polygonal approximation of digital image boundary.,2834
initial The points finest of a starts curve. technique with segmentation,The technique starts with finest initial segmentation points of a curve.,2835
images training large a Annotating time-consuming. is of number very,Annotating a large number of training images is very time-consuming.,2836
"noisy Conventionally, the looks correcting web community to samples. informative select to labels the","Conventionally, the community looks to correcting the noisy web labels to select informative samples.",2837
two of distributions the different between data. data kinds,different data distributions between the two kinds of data.,2838
then Our the registration follows fusion. method pipeline and first of,Our method follows the pipeline of first registration and then fusion.,2839
which People unknown. or frequently uncertain outcomes challenging are face problems decision-making in,People frequently face challenging decision-making problems in which outcomes are uncertain or unknown.,2840
(AI) tasks. exist intelligence algorithms can humans such learning at outperform that Artificial,Artificial intelligence (AI) algorithms exist that can outperform humans at learning such tasks.,2841
beat We isolation. both can in human team performance agent and find that performance,We find that team performance can beat both human and agent performance in isolation.,2842
uninterrupted behavior systems by sequences neuronal display Many cascading neural characterized of firing.,Many neural systems display cascading behavior characterized by uninterrupted sequences of neuronal firing.,2843
"that we networks. as Moreover, and structural in cycles such dynamic prevalent motifs observe are","Moreover, we observe cycles as structural and dynamic motifs that are prevalent in such networks.",2844
"model, this derivations of first-principles However, various is expectation. this with none among consistent","However, among various first-principles derivations of this model, none is consistent with this expectation.",2845
are derived Two assumptions. modified Hassell additional some models under,Two additional Hassell models are derived under some modified assumptions.,2846
"few learn novel just by examples. recognize contrast, a to from categories Humans, can","Humans, by contrast, can learn to recognize novel categories from just a few examples.",2847
"preventable, the leading of cause is irreversible blindness Glaucoma world-wide.","Glaucoma is the leading cause of preventable, irreversible blindness world-wide.",2848
Glaucoma recommended and is screening detection for treatment. early,Glaucoma screening is recommended for early detection and treatment.,2849
We densely aligned a re-identification person framework. propose semantically,We propose a densely semantically aligned person re-identification framework.,2850
vague other participants found to hard explanation generalize and feature instances. The to importance-based it,The participants found feature importance-based explanation vague and hard to generalize it to other instances.,2851
problem. original optimization The is into combinatorial problem a formulated,The original problem is formulated into a combinatorial optimization problem.,2852
proposed demonstrate is Experimental solutions. known results best method the to that able efficiently find,Experimental results demonstrate that the proposed method is able to efficiently find best known solutions.,2853
compared with results visual block the from statistics inspection. The analysis from were automatic,The results from the automatic analysis were compared with block statistics from visual inspection.,2854
studied genetics. widely a in population recombination Galled model as are trees,Galled trees are widely studied as a recombination model in population genetics.,2855
into relaxing galled structural networks condition. of by phylogenetic networks This is a class generalized,This class of phylogenetic networks is generalized into galled networks by relaxing a structural condition.,2856
"learn"" approach to paper figure-ground ""learning segmentation. presents image a to This","This paper presents a ""learning to learn"" approach to figure-ground image segmentation.",2857
"on issues, HMM based these (WCNN-PHMM). CNN propose a we writer-aware parsimonious To address","To address these issues, we propose a writer-aware CNN based on parsimonious HMM (WCNN-PHMM).",2858
The technique any problem. transfer applicable learning is to,The technique is applicable to any transfer learning problem.,2859
of common for authentication a verification an biometric used is Fingerprint individual. and,Fingerprint is a common biometric used for authentication and verification of an individual.,2860
and of inpainting. denoising from fingerprint extraction The a image the degraded requires,The extraction of the fingerprint from a degraded image requires denoising and inpainting.,2861
"scheme. selection state conventional First, the antenna we","First, we state the conventional antenna selection scheme.",2862
outage of terms are system secrecy secrecy probability. and in The rate simulation presented results,The simulation results are presented in terms of system secrecy rate and secrecy outage probability.,2863
among in chirp coefficients to subcarriers respect subcarriers. analyzed are cross-correlation of Orthogonal characteristics,Orthogonal characteristics of chirp subcarriers are analyzed in respect to cross-correlation coefficients among subcarriers.,2864
"Moreover, chirp systems. orthogonal developed MCDM transform implement to is (OCT)","Moreover, orthogonal chirp transform (OCT) is developed to implement MCDM systems.",2865
waveforms and systems architectures. multicarrier advantages have both chirp MCDM of Proposed,Proposed MCDM systems have both advantages of chirp waveforms and multicarrier architectures.,2866
discussed is Computational detail. proposed in detector complexity our of,Computational complexity of our proposed detector is discussed in detail.,2867
"perform they However, annotated amount with food best when large (labeled) of trained images.","However, they perform best when trained with large amount of annotated (labeled) food images.",2868
"problematic volume, impractical. obtained laborious large is expensive, are in because and when they This","This is problematic when obtained in large volume, because they are expensive, laborious and impractical.",2869
al. Goodfellow We make the training enhancements unsupervised introduced et new by to architecture,We make new enhancements to the unsupervised training architecture introduced by Goodfellow et al.,2870
paper can computed. studies be This whether rationality,This paper studies whether rationality can be computed.,2871
expected is biological by physical measured accuracy brain) of model. the or the,physical or biological brain) is measured by the expected accuracy of the model.,2872
problems. The rationality convex conclusion is optimization be that computed for can reached,The conclusion reached is that rationality can be computed for convex optimization problems.,2873
networks. propose capillary model blood emergence a the new for of We,We propose a new model for the emergence of blood capillary networks.,2874
imposed. is capillary the No between connectivity elements,No connectivity between the capillary elements is imposed.,2875
has radar. communications been communication important field an Over-water wireless of and research,Over-water communication has been an important research field of wireless communications and radar.,2876
on evaluate the Activity Collective We and datasets. method Volleyball our,We evaluate our method on the Volleyball and Collective Activity datasets.,2877
periodic In multi action hardly reaction and reality perspective situation. person are activities a in,In reality action and reaction activities are hardly periodic in a multi person perspective situation.,2878
in highly a-periodic surveillance their complex are Also for detection challenging recognizing system. gestures,Also recognizing their complex a-periodic gestures are highly challenging for detection in surveillance system.,2879
that challenging. model two operates single is these regimes Training in a simultaneously,Training a single model that operates in these two regimes simultaneously is challenging.,2880
"offers view Notably, zero-shot developing for models. new COSMO a","Notably, COSMO offers a new view for developing zero-shot models.",2881
detecting simultaneously. are They of adversarial and correctly classifying examples incapable,They are incapable of detecting and correctly classifying adversarial examples simultaneously.,2882
innovative AFG image is and for adversarial both example. an encoding,AFG is an innovative encoding for both image and adversarial example.,2883
group and consists features It label. of mixed a,It consists of group features and a mixed label.,2884
"important applications. video for (ReID) in re-identification is computer an Person vision, especially problem surveillance","Person re-identification (ReID) is an important problem in computer vision, especially for video surveillance applications.",2885
"become triplet framework a Recently, learning ReID. with common loss have for deep person networks","Recently, deep learning networks with triplet loss have become a common framework for person ReID.",2886
"on focuses triplet obtaining orders correct However, on set. loss the training","However, triplet loss focuses on obtaining correct orders on the training set.",2887
that a inferior clustering it performs task. in We demonstrate,We demonstrate that it performs inferior in a clustering task.,2888
with estimation voxel obtained from be Final can easily post-processing. then simple density map,Final estimation can then be easily obtained from voxel density map with simple post-processing.,2889
"our networks. on method the baselines and backbone two outperforms datasets In experiments, two","In experiments, our method outperforms the baselines on two datasets and two backbone networks.",2890
approach. about extensive report our We ablation also studies,We also report extensive ablation studies about our approach.,2891
successfully in Convolution applied neural image are (CNNs) task. netwotks recognition,Convolution neural netwotks (CNNs) are successfully applied in image recognition task.,2892
pitch rotation and each camera orientation video. for,camera orientation pitch and rotation for each video.,2893
the prior standard as in requires distribution Gaussian a the latent space. VAE,VAE requires the standard Gaussian distribution as a prior in the latent space.,2894
"specific this, distribution the introduces for latent paper avoid this class code. To the","To avoid this, this paper introduces the class specific distribution for the latent code.",2895
models templates. separate minutiae Two extraction provide complementary minutiae,Two separate minutiae extraction models provide complementary minutiae templates.,2896
"rare few examples activities of a interest available. events, with labeled are Many only","Many activities of interest are rare events, with only a few labeled examples available.",2897
Our more end-to-end can from model and trainable is benefit examples. few-shot,Our model is end-to-end trainable and can benefit from more few-shot examples.,2898
"the conventional indicators the time, same objective the At superior also methods. to evaluation are","At the same time, the objective evaluation indicators are also superior to the conventional methods.",2899
hypotheses spiral concept of scientific relates evolution. to paper and introduces This this the pragmatic,This paper introduces pragmatic hypotheses and relates this concept to the spiral of scientific evolution.,2900
"for two-stage and FAM. FAE theories Additionally, we of both this the underlying introduce pipeline","Additionally, we introduce the underlying theories of this two-stage pipeline for both FAE and FAM.",2901
"performance commonly in Second, analysis attribute presented. and used facial datasets metrics are the","Second, the datasets and performance metrics commonly used in facial attribute analysis are presented.",2902
"relevant attribute Furthermore, facial several issues applications. are well introduced, as related real-world as additional","Furthermore, several additional facial attribute related issues are introduced, as well as relevant real-world applications.",2903
"research promising directions. and Finally, possible we challenges future discuss","Finally, we discuss possible challenges and promising future research directions.",2904
conventions are among two types There pathologists. accepted widely of grading,There are two types of grading conventions widely accepted among pathologists.,2905
"an important hand, inter-observer the variability issue other diagnosis. for final On is","On the other hand, inter-observer variability is an important issue for final diagnosis.",2906
"image is cervical benchmark paper, this lesions precursor for grading cancer whole-slide introduced. a In","In this paper, a whole-slide image grading benchmark for cervical cancer precursor lesions is introduced.",2907
"penalize for It unacceptable equally networks to seems, classes. between missclassification however,","It seems, however, unacceptable to penalize networks equally for missclassification between classes.",2908
"representation of the is appearance quality With each RRU, improved. frame's","With RRU, the quality of each frame's appearance representation is improved.",2909
"enhance STIM. of training multi-level used RRU to capability and is the Meanwhile, objective the","Meanwhile, the multi-level training objective is used to enhance the capability of RRU and STIM.",2910
the and of distinct This faces. details to more synthesized leads shaper visual effect,This leads to shaper details and more distinct visual effect of the synthesized faces.,2911
"brings convergence In addition, the faster it during training.","In addition, it brings the faster convergence during training.",2912
the is The validated through of proposed extensive effectiveness studies. method ablation,The effectiveness of the proposed method is validated through extensive ablation studies.,2913
qualitatively our expression on and approach and also We evaluate attribute facial quantitatively facial synthesis.,We also evaluate our approach qualitatively and quantitatively on facial attribute and facial expression synthesis.,2914
of The the from optimal relay link. middle backward is location the slightly,The optimal relay location is slightly backward from the middle of the link.,2915
is when of performance zero. BER FM-EDFA The mode-dependent best gain is the,The BER performance is the best when mode-dependent gain of FM-EDFA is zero.,2916
of Reinforcement the Hierarchical realm (HRL). apparent Learning in need is particularly This,This need is particularly apparent in the realm of Hierarchical Reinforcement Learning (HRL).,2917
HRL too testbeds Goldilocks Existing exhibit often are (e.g. a simple problem; either they,Existing HRL testbeds exhibit a Goldilocks problem; they are often either too simple (e.g.,2918
Montezuma's the Arcade Learning Revenge Environment). from,Montezuma's Revenge from the Arcade Learning Environment).,2919
"optical is (using LiteFlowNet) flow features. temporal Second, to used extract","Second, optical flow (using LiteFlowNet) is used to extract temporal features.",2920
user's for propose trip novel a prediction histories. trip analyzing We approach by,We propose a novel approach for trip prediction by analyzing user's trip histories.,2921
attended yield statistically to are features These identity-discriminative representations. aggregated,These attended features are statistically aggregated to yield identity-discriminative representations.,2922
reconstructing from problem biomedical projections. the addresses work of signals This lower their dimensional,This work addresses the problem of reconstructing biomedical signals from their lower dimensional projections.,2923
Sensing for Traditionally have this (CS) employed been task. techniques based Compressed,Traditionally Compressed Sensing (CS) based techniques have been employed for this task.,2924
the of approach inductive inversion proposes a This new process. learning for work,This work proposes a new approach for inductive learning of the inversion process.,2925
Learning. is Analysis on Dictionary based Coupled It,It is based on Coupled Analysis Dictionary Learning.,2926
"intractable. of are models various statistical fields, In interest analytically","In various fields, statistical models of interest are analytically intractable.",2927
"computational is statistical by a inference result, hampered constraints. greatly As","As a result, statistical inference is greatly hampered by computational constraints.",2928
"likely given similar data users perform are different However, computations. to with model, different a","However, given a model, different users with different data are likely to perform similar computations.",2929
Computations other user for users potentially sets. one by data with are useful different done,Computations done by one user are potentially useful for other users with different data sets.,2930
to of capitalize a this. pooling on across resources researchers We propose,We propose a pooling of resources across researchers to capitalize on this.,2931
"techniques, any solved now problem Using on interpolation estimation be advanced the can individual spot.","Using advanced interpolation techniques, any individual estimation problem can now be solved on the spot.",2932
parameters. can the The easily method on prepaid constraints priors well as different accommodate as,The prepaid method can easily accommodate different priors as well as constraints on the parameters.,2933
The evaluated is performance Leave-One-Subject-Out cross by validation.,The performance is evaluated by Leave-One-Subject-Out cross validation.,2934
"is segmenting However, time-consuming of and labor-intensive. NPC manually","However, manually segmenting of NPC is time-consuming and labor-intensive.",2935
"enough Additionally, generally information MRI delineation. accurate its provide cannot for single-modality","Additionally, single-modality MRI generally cannot provide enough information for its accurate delineation.",2936
A fuse presented features MRI. block to multi-modality fusion from is effectively,A fusion block is presented to effectively fuse features from multi-modality MRI.,2937
"are several degradation. in mixture of a of often practice degradations types image However,","However, image degradations in practice are often a mixture of several types of degradation.",2938
The source ill-posed localization EEG an problem. is,The EEG source localization is an ill-posed problem.,2939
which the of of involves estimation measurements. outnumbers the sources It number,It involves estimation of the sources which outnumbers the number of measurements.,2940
This approach for presents EEG dense localization. source array new a paper,This paper presents a new approach for dense array EEG source localization.,2941
trainable the end-to-end problem. for We a graph framework propose novel decomposition,We propose a novel end-to-end trainable framework for the graph decomposition problem.,2942
"neural convolutional image good Recently, recognition. networks deep results have shown for","Recently, deep convolutional neural networks have shown good results for image recognition.",2943
to our We of the method propose recognition for analysis. crystals protein structural X-ray applying,We propose applying our method to the recognition of protein crystals for X-ray structural analysis.,2944
networks. methods protein neural several recognition realize by crystallization using convolutional that are There,There are several methods that realize protein crystallization recognition by using convolutional neural networks.,2945
"to accuracy. are data In high method, with required recognize large-scale sets each","In each method, large-scale data sets are required to recognize with high accuracy.",2946
"of CNN. data training our the not for set, images enough good number In is","In our data set, the number of images is not good enough for training CNN.",2947
serious is for amount data fields. a of various issue in CNN The,The amount of data for CNN is a serious issue in various fields.,2948
crystallization several our We is method effective demonstrate for images experiments. that using,We demonstrate that our method is effective for crystallization images using several experiments.,2949
network several tune. architectures complicated This leads to to introduced hyperparameters with,This leads to complicated network architectures with several introduced hyperparameters to tune.,2950
object networks detection region location. (RPN) hypothesize object methods to use State-of-the-art for proposal,State-of-the-art methods for object detection use region proposal networks (RPN) to hypothesize object location.,2951
that Recent Background: during rest. even have dynamic indicated studies functional is connectivity,Background: Recent studies have indicated that functional connectivity is dynamic even during rest.,2952
the females estimated The is in dynamic between difference connectivity states males and identified.,The difference in the estimated dynamic connectivity states between males and females is identified.,2953
"the function. by IVQR and complicated objective estimation the of GMM However, non-convexity is non-smoothness","However, estimation is complicated by the non-smoothness and non-convexity of the IVQR GMM objective function.",2954
"tries it, (\textit{e.g. one to address exploit To strategies group mining-based","To address it, one group tries to exploit mining-based strategies (\textit{e.g.",2955
"examples. mining hard focal loss) informative focus }, to example and on the","}, hard example mining and focal loss) to focus on the informative examples.",2956
designing margin-based functions The loss (\textit{e.g. group other devotes to,The other group devotes to designing margin-based loss functions (\textit{e.g.,2957
discriminative them of features. learn Both have to well-verified been,Both of them have been well-verified to learn discriminative features.,2958
state-of-the-arts. over approach effectiveness several benchmarks the on of our have Experimental results demonstrated,Experimental results on several benchmarks have demonstrated the effectiveness of our approach over state-of-the-arts.,2959
human an untrimmed in important task. actions is Recognizing challenging videos,Recognizing human actions in untrimmed videos is an important challenging task.,2960
of observations. requires radar global and nowcasting precipitation Truly satellite fusion,Truly global precipitation nowcasting requires fusion of radar and satellite observations.,2961
these to tackle retinopathy of problems. essential Automated diabetic detection is,Automated detection of diabetic retinopathy is essential to tackle these problems.,2962
models Results show that is DeepBillboard various effective for scenes. steering and,Results show that DeepBillboard is effective for various steering models and scenes.,2963
are results of summarized The in analysis the sheet. tally the,The results of the analysis are summarized in the tally sheet.,2964
view. a surgery limited of field Laparoscopic has,Laparoscopic surgery has a limited field of view.,2965
"Laser smoke, laproscopic in a influences surgery causes inevitably surgeon's ablation which visibility. the","Laser ablation in a laproscopic surgery causes smoke, which inevitably influences the surgeon's visibility.",2966
utilizes proposed method framework~(WLS). weighted squares least Our optimization,Our proposed method utilizes weighted least squares optimization framework~(WLS).,2967
SVM classification a smoke/non-smoke We task. then train binary with classifier,We then train a SVM classifier with binary smoke/non-smoke classification task.,2968
"greatly and in understanding, help interpreting Images visualizing data.","Images greatly help in understanding, interpreting and visualizing data.",2969
the accessibility. Adding of images principle web foremost description and to textual is first,Adding textual description to images is the first and foremost principle of web accessibility.,2970
"image algorithms processing information relevant from images. uses to specific extract it the chart Finally,","Finally, it uses specific image processing algorithms to extract relevant information from the chart images.",2971
"ViewNet, deep we DispairtyNet networks three - these neural propose RefineNet. three For and stages","For these three stages we propose three deep neural networks - ViewNet, DispairtyNet and RefineNet.",2972
to and noise cannot the precisely. superpixel are Most methods superpixel sensitive control number,Most superpixel methods are sensitive to noise and cannot control the superpixel number precisely.,2973
fast onion algorithm and a precise (OP) control superpixel We called number peeling algorithm. develop,We develop a fast and precise superpixel number control algorithm called onion peeling (OP) algorithm.,2974
without much superpixel the number The control computational reducing accurately can algorithm efficiency. OP,The OP algorithm can control the superpixel number accurately without reducing much computational efficiency.,2975
area. and geometric research deep fruitful learning for is active data Developing an techniques,Developing deep learning techniques for geometric data is an active and fruitful research area.,2976
representation. of sphere-type This novel surface the tackles surface-to-image problem a learning by paper developing,This paper tackles the problem of sphere-type surface learning by developing a novel surface-to-image representation.,2977
Real-Time games a have artificial Strategy intelligence popular (RTS) recently testbed research. for become,Real-Time Strategy (RTS) games have recently become a popular testbed for artificial intelligence research.,2978
complex number represent AI of challenges. providing domain a adversarial interesting a They,They represent a complex adversarial domain providing a number of interesting AI challenges.,2979
"algorithms, the usage. their and compare them, We both demonstrate evaluate","We evaluate both the algorithms, compare them, and demonstrate their usage.",2980
refer (SMPLR). as to SMPL reverse We this,We refer to this as SMPL reverse (SMPLR).,2981
Person the single-domain re-identification under significant has achieved (ReID) setting. improvement,Person re-identification (ReID) has achieved significant improvement under the single-domain setting.,2982
"brings First, (PAP) improvement that Pooling significant testing. for Aligned we cross-domain propose Part","First, we propose Part Aligned Pooling (PAP) that brings significant improvement for cross-domain testing.",2983
both and state-of-the-art performance achieves under settings. Our cross-domain source-domain model,Our model achieves state-of-the-art performance under both source-domain and cross-domain settings.,2984
"For our completeness, we model also domain existing complementarity adaptation of demonstrate to the methods.","For completeness, we also demonstrate the complementarity of our model to existing domain adaptation methods.",2985
optimization problems. for proposed solutions subsequently are the corresponding Tractable,Tractable solutions are proposed subsequently for the corresponding optimization problems.,2986
"breast brain tissues, human imaging: from mammalian human images tissues. tissues, and Medical histopathological acquired","Medical imaging: histopathological images acquired from mammalian tissues, human breast tissues, and human brain tissues.",2987
"face, dogs, classification: object objects, flowers, General etc. indoor scenes,","General object classification: face, flowers, objects, dogs, indoor scenes, etc.",2988
furnished The with limit describing theoretical presented and memory results algorithm is asymptotic its usage.,The presented algorithm is furnished with theoretical results describing its asymptotic limit and memory usage.,2989
life-long in far been small-scale learning setups. studied relatively and So has (LLL) artificial relatively,So far life-long learning (LLL) has been studied in relatively small-scale and relatively artificial setups.,2990
"we Here, introduce new a large-scale alternative.","Here, we introduce a new large-scale alternative.",2991
"<person, wave> Third, tasks <dog, structure (e.g., riding, and share facts wave>). riding, across may","Third, facts across tasks may share structure (e.g., <person, riding, wave> and <dog, riding, wave>).",2992
"of facts, LLL Given large possible setup a seems natural choice. the a number","Given the large number of possible facts, a LLL setup seems a natural choice.",2993
and for method between domain Our domains. accounts within explicitly stylistic shifts,Our method explicitly accounts for stylistic domain shifts between and within domains.,2994
"generating we value. find such while economic POMDP-based contrast, management higher avoids In over-exploitation also","In contrast, we find POMDP-based management avoids such over-exploitation while also generating higher economic value.",2995
R method. implements an Includes function which the,Includes an R function which implements the method.,2996
superior FCN. Extensive validate to MCPNet like that DNNs is the experiments conventional,Extensive experiments validate that the MCPNet is superior to conventional DNNs like FCN.,2997
from are features The image domains. two extracted,The image features are extracted from two domains.,2998
adaptive based on sampling algorithm image Learning presents paper an (DL). This Deep,This paper presents an adaptive image sampling algorithm based on Deep Learning (DL).,2999
inpainting network. mask with sampling is network generation image adaptive an jointly The trained,The adaptive sampling mask generation network is jointly trained with an image inpainting network.,3000
captioning task. for hLSTMat video our We initially design,We initially design our hLSTMat for video captioning task.,3001
"it to we further task. captioning and it Then, apply refine image","Then, we further refine it and apply it to image captioning task.",3002
in components study. also The exploited ablation is important of well the effect,The effect of important components is also well exploited in the ablation study.,3003
"Finally, the architecture. proposed demonstrate results numerical also efficiency of the robustness, effectiveness, and","Finally, the numerical results also demonstrate effectiveness, robustness, and efficiency of the proposed architecture.",3004
"them to identify as appear. and as Therefore, soon critical it they detect is","Therefore, it is critical to detect and identify them as soon as they appear.",3005
in-painter with is built patch discriminators. coarse-to-fine framework (generator) and two The class-aware a,The framework is built with a coarse-to-fine patch in-painter (generator) and two class-aware discriminators.,3006
by of computational pixel resources. requires amount enormous pixel classical Representing using image information an,Representing image pixel by pixel using classical information requires an enormous amount of computational resources.,3007
"information is a images paradigm of important. in represent Hence, methods to exploring different","Hence, exploring methods to represent images in a different paradigm of information is important.",3008
"the In Quantum representation this work, Information. images we in of study","In this work, we study the representation of images in Quantum Information.",3009
"clustering proposed In on (NS) based new neutrosophic this theory. algorithm work, a set is","In this work, a new clustering algorithm is proposed based on neutrosophic set (NS) theory.",3010
versa. data is regions points Lower dense indeterminacy to and in assigned vice,Lower indeterminacy is assigned to data points in dense regions and vice versa.,3011
"cost In by methods. is function minimized third descend proposed the the step, gradient","In the third step, the proposed cost function is minimized by gradient descend methods.",3012
based on membership are their degrees. Data points clustered,Data points are clustered based on their membership degrees.,3013
space. for Video holds and many utility labelling the data applications processing in emerging IoT,Video processing holds utility for many emerging applications and data labelling in the IoT space.,3014
"small have kinase-targeted been FDA drugs approved. molecule Currently, EGFR of generations three","Currently, three generations of EGFR kinase-targeted small molecule drugs have been FDA approved.",3015
"acquired However, and treatment long-term to drug results resistance in NSCLC. further vulnerability","However, long-term treatment results in acquired drug resistance and further vulnerability to NSCLC.",3016
"needed. overcome kinase Therefore, are specially novel that urgently EGFR inhibitors mutations acquired","Therefore, novel EGFR kinase inhibitors that specially overcome acquired mutations are urgently needed.",3017
"viewpoint, organization in this space. identity of gender, We the and illumination, examined","We examined the organization of viewpoint, illumination, gender, and identity in this space.",3018
Deep long-standing problems networks recognition. face generalized computational in solve produce face representations that,Deep networks produce face representations that solve long-standing computational problems in generalized face recognition.,3019
develop $K$-fold bias-correction. a procedure for We cross-fitting,We develop a $K$-fold cross-fitting procedure for bias-correction.,3020
is R-package An available. methods our implementing for,An R-package for implementing our methods is available.,3021
processing. sensing is Hyperspectral remote a in research data topic prominent,Hyperspectral remote sensing is a prominent research topic in data processing.,3022
linear unmixing spectral adopting models. by the Most algorithms developed mixing of are the,Most of the spectral unmixing algorithms are developed by adopting the linear mixing models.,3023
"NMF, Sparsity constraints regularized L_ to $ was $ added and {q} by norm. was","Sparsity constraints was added to NMF, and was regularized by $ L_ {q} $ norm.",3024
"In proposed the including algorithm, a employed. network is clusters","In the proposed algorithm, a network including clusters is employed.",3025
as the images node network. hyperspectral a in this pixel considered in Each,Each pixel in the hyperspectral images considered as a node in this network.,3026
objective such functions common the Two is sum-of-costs the makespan. and,Two such common objective functions is the sum-of-costs and the makespan.,3027
the of we in study for compilation-based applicability solvers Moreover approach. developed search-based techniques various,Moreover we study applicability of various techniques developed for search-based solvers in the compilation-based approach.,3028
"We borrowing by solver. improve further the a ICTS, encoding ideas search-based then from","We then further improve the encoding by borrowing ideas from ICTS, a search-based solver.",3029
mesh generate skinned to This allows turn in us animations.,This in turn allows us to generate skinned mesh animations.,3030
"we regression approach. two-step a Here, propose","Here, we propose a two-step regression approach.",3031
"our phenomena in deception in lives. are both private and and society, common professional Lies","Lies and deception are common phenomena in society, both in our private and professional lives.",3032
"notoriously are deception accurate However, humans at detection. bad","However, humans are notoriously bad at accurate deception detection.",3033
of in the makes salt presence computer-aided dome The more data interpretation noise even challenging.,The presence of noise in the data makes computer-aided salt dome interpretation even more challenging.,3034
"detection. dome we a utilize salt accomplish gradient to texture-based First,","First, we utilize a texture-based gradient to accomplish salt dome detection.",3035
"noise. algorithm shows the is that At time, analysis the our same proposed to robust","At the same time, our analysis shows that the proposed algorithm is robust to noise.",3036
"Compared defenses, advantages. baseline has with three DUP-Net","Compared with baseline defenses, DUP-Net has three advantages.",3037
"denoising it is added statistical since robustness the outlier removal a Second, provides operation. non-differentiable","Second, the statistical outlier removal provides added robustness since it is a non-differentiable denoising operation.",3038
that DUP-Net We validate experiments conduct as very is effective various defense practice. in to,We conduct various experiments to validate that DUP-Net is very effective as defense in practice.,3039
"optimize representation. double regret neural algorithm the counterfactual Furthermore, adopt this to we minimization","Furthermore, we adopt the counterfactual regret minimization algorithm to optimize this double neural representation.",3040
network on generates is fully learned convolutional network based residual deeply Our features. that,Our network is based on fully convolutional network that generates deeply learned residual features.,3041
We design discriminative function a soft-shifted triplet to palmprint features. more learn loss effectively,We design a soft-shifted triplet loss function to more effectively learn discriminative palmprint features.,3042
But ideal hold? do these if not what conditions,But what if these ideal conditions do not hold?,3043
"required analyses, to it efficient against contaminations. robust use more For is estimators","For more efficient analyses, it is required to use robust estimators against contaminations.",3044
convolutional exploit layers. fully from hierarchical We features all the the,We fully exploit the hierarchical features from all the convolutional layers.,3045
"accurate salient fusion should strategies complex However, and the be object more more for detection?","However, should the fusion strategies be more and more complex for accurate salient object detection?",3046
"contextual estimation. At last, used for can hybrid features accurate be saliency the","At last, the hybrid contextual features can be used for accurate saliency estimation.",3047
paper upon will released Code acceptance. be,Code will be released upon paper acceptance.,3048
benchmark The are underidentified. preference and jointly parameters DM's model,The DM's benchmark model and preference parameters are jointly underidentified.,3049
allows rates of to be characterized. result also estimators The convergence,The result also allows convergence rates of estimators to be characterized.,3050
the pessimistically distorts reveals benchmark application in time-variation DM way the The probabilities.,The application reveals time-variation in the way the DM pessimistically distorts benchmark probabilities.,3051
"To considered knowledge, have not best methods this the fact. our IQA previous of","To the best of our knowledge, previous IQA methods have not considered this fact.",3052
stronger values. method comparison correlation of the our reveals with human-evaluated This,This comparison reveals the stronger correlation of our method with human-evaluated values.,3053
consistent independent cases. distortion for prediction-of-quality well is specific as score The as distortion,The prediction-of-quality score is consistent for distortion specific as well as distortion independent cases.,3054
"Moreover, makes faster to any application. processing real-time it applicable","Moreover, faster processing makes it applicable to any real-time application.",3055
systems the characteristic electrical distribution (EDS). most against is major essential of disasters Resilience future,Resilience against major disasters is the most essential characteristic of future electrical distribution systems (EDS).,3056
this method for restoration proposed EDS paper. A is rolling optimization multi-agent-based in scheduling,A multi-agent-based rolling optimization method for EDS restoration scheduling is proposed in this paper.,3057
is for entire A rolling restoration EDS the proposed. optimization process,A rolling optimization process for the entire EDS restoration is proposed.,3058
to problem solve trend globally this current the is framework. BnB optimization The using,The current trend is to solve this problem globally using the BnB optimization framework.,3059
"techniques propose these address to this In two paper, problems. we","In this paper, we propose two techniques to address these problems.",3060
and decomposition algorithms substantially the BnB its dimensionality improves transformation This of efficiency. problem reduces,This transformation decomposition reduces the problem dimensionality of BnB algorithms and substantially improves its efficiency.,3061
bounding tightly-fitted yields boxes process or that This occlusions circumvent miss-detections.,This process yields tightly-fitted bounding boxes that circumvent occlusions or miss-detections.,3062
efficiency without we strive simplicity accuracy. for architecture recognition In proposed sacrificing our and,In our proposed architecture we strive for simplicity and efficiency without sacrificing recognition accuracy.,3063
"illumination performance descriptors such severe drops However, of under changes. the","However, the performance of such descriptors drops under severe illumination changes.",3064
"In local and a Invertible descriptor. Discriminative proposed this paper, Contrast (DCI) we feature","In this paper, we proposed a Discriminative and Contrast Invertible (DCI) local feature descriptor.",3065
is based estimate A on region. a divergence of local proposed the flipping robust contrast,A robust contrast flipping estimate is proposed based on the divergence of a local region.,3066
"we identify user (attributes) survey. to cohesion, order the conducted a visual In for cues","In order to identify the visual cues (attributes) for cohesion, we conducted a user survey.",3067
via multi-task at neural a is group-level analysis a performed network. Image convolutional,Image analysis is performed at a group-level via a multi-task convolutional neural network.,3068
GCS and the `GAF-Cohesion to database add the We database'. Affect propose Group,We add GCS to the Group Affect database and propose the `GAF-Cohesion database'.,3069
cohesion that group-level and are This suggests emotion correlated.,This suggests that group-level emotion and cohesion are correlated.,3070
of Face resolution low noise analysis for applications and face recognition. important from is restoration,Face restoration from low resolution and noise is important for applications of face analysis recognition.,3071
data. also the of labeled in input They absence suffer,They also suffer in the absence of labeled input data.,3072
laziness. This dual productive concept be objective the of summarized can by,This dual objective can be summarized by the concept of productive laziness.,3073
overlook worker on scheduling rest. wellbeing Existing proper focus efficiency approaches but mostly through,Existing scheduling approaches mostly focus on efficiency but overlook worker wellbeing through proper rest.,3074
Our by is Algorithm (LISTA). inspired the Learned Shrinkage-Threshold Iterative model,Our model is inspired by the Learned Iterative Shrinkage-Threshold Algorithm (LISTA).,3075
method. Extensive the our benchmark on demonstrate effectiveness experiments of datasets,Extensive experiments on benchmark datasets demonstrate the effectiveness of our method.,3076
"dramatically. In past demand services years, data the for increased few has high rate the","In the past few years, the demand for high data rate services has increased dramatically.",3077
"dispersion systems mobility mitigation and we modelling, techniques. describe indoor In OWC channel","In indoor OWC systems we describe channel modelling, mobility and dispersion mitigation techniques.",3078
"research open in systems OWC Moreover, issues are discussed.","Moreover, open research issues in OWC systems are discussed.",3079
transform. realized wavelet by The transformation fast a is,The transformation is realized by a fast wavelet transform.,3080
model has used classification for image Knowledge (KD) Distillation been in compression.,Knowledge Distillation (KD) has been used in image classification for model compression.,3081
"However, on studies apply technology rare detectors. object single-stage this","However, rare studies apply this technology on single-stage object detectors.",3082
encountered the in This when applying also task. is problem detection KD,This problem is also encountered when applying KD in the detection task.,3083
"than others. hard teacher-defined the far samples important more For KD, are any","For KD, the teacher-defined hard samples are far more important than any others.",3084
"inverted index memory effectively. However, requires high index the dataset overhead the to","However, the inverted index requires high memory overhead to index the dataset effectively.",3085
"consumption. comparable maintaining while both search method accuracy, Our quantization and improves efficiency memory","Our quantization method improves both search efficiency and accuracy, while maintaining comparable memory consumption.",3086
regions. by number search indexed and is reducing the increasing This of space achieved,This is achieved by reducing search space and increasing the number of indexed regions.,3087
"further its performance. limited However, information improving spectral classification the hinders","However, its limited spectral information hinders further improving the classification performance.",3088
the of also proposed system adaptive discussed. control Practical is implementation polarization with,Practical implementation of the proposed system with adaptive polarization control is also discussed.,3089
estimation solutions blurry produce depth often low approximations resolution. of Existing for,Existing solutions for depth estimation often produce blurry approximations of low resolution.,3090
publicly Code corresponding available. pre-trained made and weights are,Code and corresponding pre-trained weights are made publicly available.,3091
"First-order is tool statistics, computer a ubiquitous across differentiation and learning, automatic machine science.","First-order automatic differentiation is a ubiquitous tool across statistics, machine learning, and computer science.",3092
"realize yet Higher-order the automatic differentiation, implementations have of same utility. to however,","Higher-order implementations of automatic differentiation, however, have yet to realize the same utility.",3093
"between In pose non-overlapping approach even scans. particular, relative estimates our encouraging provides","In particular, our approach provides encouraging relative pose estimates even between non-overlapping scans.",3094
the CNN two are from which tasks feature share extracted. The multi-scale maps a backbone,The two tasks share a backbone CNN from which the multi-scale feature maps are extracted.,3095
a for flexible This or occluded keypoints. creates truncated localizing representation,This creates a flexible representation for localizing occluded or truncated keypoints.,3096
its the does circuit learning of a size influence neural How performance?,How does the size of a neural circuit influence its learning performance?,3097
in found Larger with tend species ability. learning and be cognitive function brains to higher,Larger brains tend to be found in species with higher cognitive function and learning ability.,3098
for objects. Computer and detecting vision relies on labeled datasets in and evaluation training recognizing,Computer vision relies on labeled datasets for training and evaluation in detecting and recognizing objects.,3099
evaluate effort a applications. large-scale to describes for novel This YOLO's consistency paper,This paper describes a novel effort to evaluate YOLO's consistency for large-scale applications.,3100
the technology Regularization is into of design architecture. integrated deep,Regularization technology is integrated into the design of deep architecture.,3101
the training is end-to-end out loss. carried The Euclidean using joint,The end-to-end joint training is carried out using the Euclidean loss.,3102
bodies generative for silhouette-based models. human introduce using representation a We modeling clothed deep new,We introduce a new silhouette-based representation for modeling clothed human bodies using deep generative models.,3103
many in communication essential removal received (IN) noise signal of impulsive from applications. Efficient is,Efficient removal of impulsive noise (IN) from received signal is essential in many communication applications.,3104
the absolute This efficient camera shutter problem. to pose paper rolling new solutions presents,This paper presents new efficient solutions to the rolling shutter camera absolute pose problem.,3105
more of to types further our We make various rooftops solution evaluated robust.,We further evaluated various types of rooftops to make our solution more robust.,3106
are by majority The mobile taken vast photos phones. today of,The vast majority of photos taken today are by mobile phones.,3107
enhance computationally these to This us motivates images.,This motivates us to computationally enhance these images.,3108
"image Over yielded last deep few learning the have improvements techniques in years, significant inpainting.","Over the last few years, deep learning techniques have yielded significant improvements in image inpainting.",3109
"each from global reciprocal features attribute-category benefit other. local this and process, Through","Through this attribute-category reciprocal process, local and global features benefit from each other.",3110
"using the images Finally, points. identified two by are together stitched","Finally, two images are stitched together by using the identified points.",3111
"Therefore, methods. unsupervised work began depth to estimation researchers on","Therefore, researchers began to work on unsupervised depth estimation methods.",3112
during together the two test models network are The phase. piped,The two network models are piped together during the test phase.,3113
We multifocal a framework propose new computing-inspired cancer procedure detection (MCDP). of,We propose a new framework of computing-inspired multifocal cancer detection procedure (MCDP).,3114
computer Convolutional primary the neural algorithms become various networks(CNNs) of vision for has tasks. one,Convolutional neural networks(CNNs) has become one of the primary algorithms for various computer vision tasks.,3115
example is task has such that of character a typical attracted Handwritten attention. recognition also,Handwritten character recognition is a typical example of such task that has also attracted attention.,3116
"that problems ancient propose a We, novel solution with for processing. therefore, the painting come","We, therefore, propose a novel solution for the problems that come with ancient painting processing.",3117
images. transfer to convert use ancient to domain This natural paintings to photo-realistic is,This is to use domain transfer to convert ancient paintings to photo-realistic natural images.,3118
"in work. on Specifically, and paintings this we bird Chinese ancient landscape flower, focus","Specifically, we focus on Chinese ancient flower, bird and landscape paintings in this work.",3119
"control many there is limited data challenges since it acquisition However, in faces conditions.","However, it faces many challenges since there is limited control in data acquisition conditions.",3120
to model released Our will this be problem. based facilitate on further studies,Our model will be released to facilitate further studies based on this problem.,3121
into pose categories: and multi-stage fall Existing estimation methods. two single-stage approaches,Existing pose estimation approaches fall into two categories: single-stage and multi-stage methods.,3122
research. The available publicly for further code is source,The source code is publicly available for further research.,3123
very The salt-dome delineation precision workflow proposed with boundary detects accuracy. good the and,The proposed delineation workflow detects the salt-dome boundary with very good precision and accuracy.,3124
Interest are this blob significance points measurement. extracted from,Interest points are extracted from this blob significance measurement.,3125
"result, is As detector changes. proposed invariant the illumination to a","As a result, the proposed detector is invariant to illumination changes.",3126
"the problem not them did of of address network many interpretability. However,","However, many of them did not address the problem of network interpretability.",3127
outperforms Our residual solution recent approaches. even learning,Our solution outperforms even recent residual learning approaches.,3128
information surrounding actions. for understanding essential The provides context,The surrounding context provides essential information for understanding actions.,3129
"leverage For which actor localization, better. pre-trained we transfer detectors, object","For actor localization, we leverage pre-trained object detectors, which transfer better.",3130
real-time and proposed achieves model pipeline our detection action near performance. efficient The is,The proposed model is efficient and our action detection pipeline achieves near real-time performance.,3131
approach an to named (CHR) refinement We propose class-balanced deal with hierarchical difficulties. these,We propose an approach named class-balanced hierarchical refinement (CHR) to deal with these difficulties.,3132
with evaluate different ratios of on samples. We CHR positive/negative SIXray,We evaluate CHR on SIXray with different ratios of positive/negative samples.,3133
"by spontaneous, seizures. characterised   disorder Epilepsy is recurrent a","  Epilepsy is a disorder characterised by spontaneous, recurrent seizures.",3134
neural success (DNNs) numerous networks extraordinary Deep achieved have areas. in,Deep neural networks (DNNs) have achieved extraordinary success in numerous areas.,3135
are happen the likely Overfitting to insufficient. such training in also is network when data,Overfitting is also likely to happen in such network when the training data are insufficient.,3136
shortcomings in DNNs platforms. application of resource-constrained hinder These the severely,These shortcomings severely hinder the application of DNNs in resource-constrained platforms.,3137
"auxiliary Besides, an employed as unnecessary to is remove sparsity group neurons. also","Besides, group sparsity is also employed as an auxiliary to remove unnecessary neurons.",3138
datasets several on of effectiveness demonstrate method. Experiments proposed the the public,Experiments on several public datasets demonstrate the effectiveness of the proposed method.,3139
"and stores. in milk juice grocery cartons, packages","milk packages and juice cartons, in grocery stores.",3140
work addresses the under semantic This of fog. problem understanding scene,This work addresses the problem of semantic scene understanding under fog.,3141
datasets at project and the The are available code website.,The datasets and code are available at the project website.,3142
and Surface-based provides object cues analysis for geometric strong modeling. semantic topology geodesic,Surface-based geodesic topology provides strong cues for object semantic analysis and geometric modeling.,3143
"lost point However, information in clouds. is connectivity such","However, such connectivity information is lost in point clouds.",3144
"the Furthermore, process. statistical properties simulations are of jump underlying the used to describe","Furthermore, the simulations are used to describe statistical properties of the underlying jump process.",3145
regularization we framework multi-task Here propose segmentation. a for MR cardiac image learning (MTL)-based,Here we propose a multi-task learning (MTL)-based regularization framework for cardiac MR image segmentation.,3146
at outperform artificial resolution networks existing neural scans. both and medium Our methods low,Our artificial neural networks outperform existing methods at both medium and low resolution scans.,3147
in computer learning Supervised based techniques state-of-the-art hungry. are data general vision machine,Supervised machine learning based state-of-the-art computer vision techniques are in general data hungry.,3148
strong function the the can mitigate One placing structural by estimation. in restrictions on problem,One can mitigate the problem by placing strong restrictions on the structural function in estimation.,3149
"does not imparts them function if the then obey However, the bias. true imposing restrictions","However, if the true function does not obey the restrictions then imposing them imparts bias.",3150
to our methods We al. Blundell setting the of empirical et apply,We apply our methods to the empirical setting of Blundell et al.,3151
terms in techniques well-known of latter outperforms estimation and robustness. accuracy state-of-the-art most The,The latter outperforms most well-known state-of-the-art techniques in terms of estimation accuracy and robustness.,3152
acid. dihedral and $\phi$ bivariate make The amino between angles $\psi$ within models an relationships,The bivariate models make relationships between dihedral angles $\phi$ and $\psi$ within an amino acid.,3153
"Furthermore, EDA we the with compared approaches literature. from other proposed the","Furthermore, we compared the proposed EDA with other approaches from the literature.",3154
capabilities increase segmentation proposed techniques. should of the varied generalization training This data the,This varied training data should increase the generalization capabilities of the proposed segmentation techniques.,3155
training update an for design an end-to-end CAN alternating We strategy manner. in,We design an alternating update strategy for training CAN in an end-to-end manner.,3156
security systems. is of recognition Face face for the crucial anti-spoofing,Face anti-spoofing is crucial for the security of face recognition systems.,3157
methods based deep need large-scale samples learning based methods reduce to overfitting. training Learning especially,Learning based methods especially deep learning based methods need large-scale training samples to reduce overfitting.,3158
"proposed face solutions presented. of recognition taxonomy, is Following comprehensive a representative the survey","Following the proposed taxonomy, a comprehensive survey of representative face recognition solutions is presented.",3159
(SNRs) of under ratios is performance SVR evaluated The experimentally. different signal-to-noise,The performance of SVR is evaluated under different signal-to-noise ratios (SNRs) experimentally.,3160
the Our a segmentation suggested performance. proposed method that results superior had,Our results suggested that the proposed method had a superior segmentation performance.,3161
image. network attributes pre-trained is to VGG-Face A used visible the from the extract,A pre-trained VGG-Face network is used to extract the attributes from the visible image.,3162
a system. the of tracking used results monopulse for be can optimizing design The,The results can be used for optimizing the design of a monopulse tracking system.,3163
a new motion learning approach from We to propose self-supervised feature image cue.,We propose a new self-supervised approach to image feature learning from motion cue.,3164
an segmentation evaluate We and demonstrate image this on task. approach,We demonstrate and evaluate this approach on an image segmentation task.,3165
addresses in videos a of manner. automatic summarization This unified paper,This paper addresses automatic summarization of videos in a unified manner.,3166
details We modalities involved. provide different implementation summarization the on systems also and,We also provide implementation details on summarization systems and the different modalities involved.,3167
musical talent and finally to be to studies performed improve propose more We abilities. musical,We finally propose more studies to be performed to improve musical talent and musical abilities.,3168
uncertainty in perspective brings Constraint new Programming. to This tackle,This brings new perspective to tackle uncertainty in Constraint Programming.,3169
remote indispensable clouds image is step pre-processing Removing analysis. an in sensing,Removing clouds is an indispensable pre-processing step in remote sensing image analysis.,3170
"applied However, rarely learning is in image removal clouds. deep remote sensing","However, deep learning is rarely applied in remote sensing image removal clouds.",3171
for of the sets lack The neural training reason is networks. data,The reason is the lack of data sets for training neural networks.,3172
has in been the protein-folding long-standing challenge a biology. and problem understanding Exploring molecular,Exploring and understanding the protein-folding problem has been a long-standing challenge in molecular biology.,3173
the The is the method from by results literature. experiments comparing in verified available,The method is verified by comparing the results available from experiments in the literature.,3174
led to these has severely direct The models problems. optimization of use ill-posed,The direct use of these models has led to severely ill-posed optimization problems.,3175
real Simulation effectiveness the data of strategy. proposed the swith synthetic attest and,Simulation swith synthetic and real data attest the effectiveness of the proposed strategy.,3176
simple a game-play. real-time game with rules is strategy but complex Wars Planet,Planet Wars is a real-time strategy game with simple rules but complex game-play.,3177
pre-segmentation on This a a based graph without framework region adjacency stage. is algorithm,This algorithm is based on a region adjacency graph framework without a pre-segmentation stage.,3178
image functions with are defined noise. normal weight edge linear from Proposed model,Proposed edge weight functions are defined from linear image model with normal noise.,3179
dataset every image. with colour provided for pixel-by-pixel The truth ground segmentation is,The dataset is provided with pixel-by-pixel ground truth colour segmentation for every image.,3180
task of vision in correspondences processing Establishing image variety a computer fundamental and is applications.,Establishing correspondences is a fundamental task in variety of image processing and computer vision applications.,3181
"LAT In for feature addition, several are with robust reformulated practical descriptors applications.","In addition, robust feature descriptors are reformulated with LAT for several practical applications.",3182
to mean as compared LAT absolute methods. reduces the conventional difference,LAT reduces the mean absolute difference as compared to conventional methods.,3183
"MDPs. of we In convergence heuristics new to this the speed paper, solution up present","In this paper, we present new heuristics to speed up the solution convergence of MDPs.",3184
this technique tackle as a problem. to recently emerged has Deep promising learning,Deep learning has recently emerged as a promising technique to tackle this problem.,3185
relative simultaneously. predict and multi-task perform We and poses absolute learning the,We perform multi-task learning and predict the absolute and relative poses simultaneously.,3186
cast problem art Prior as often detection. this boundary,Prior art often cast this problem as boundary detection.,3187
Our and previous quantitatively surpasses methods method qualitatively.,Our method surpasses previous methods quantitatively and qualitatively.,3188
matching of across is the problem objects cameras. appearances of Re-identification different,Re-identification is the problem of matching appearances of objects across different cameras.,3189
"include arising smart occlusions. viewpoints, city scenarios variations illumination Typical and of in challenges self","Typical challenges arising in smart city scenarios include variations of viewpoints, illumination and self occlusions.",3190
Popular embedding triplet functions contrastive loss loss. (space) learning an or include for,Popular loss functions for learning an embedding (space) include contrastive or triplet loss.,3191
hierarchical structure video recurrent to of network Our uses temporal capture approach a the features.,Our approach uses a hierarchical recurrent network to capture the temporal structure of video features.,3192
from appearance networks. detect motion features and accidents by two-stream We incorporating features near,We detect near accidents by incorporating appearance features and motion features from two-stream networks.,3193
use strategy model. to We the the method alphaGo explain of gaming Zero our,We use our method to explain the gaming strategy of the alphaGo Zero model.,3194
analysis contextual local of The w.r.t. effects,The analysis of local contextual effects w.r.t.,3195
special units is real values input in applications. certain of,certain input units is of special values in real applications.,3196
Explaining alphaGo logic a is Zero model application. the of the typical,Explaining the logic of the alphaGo Zero model is a typical application.,3197
"In of the method move our during experiments, rationale successfully the game. Go disentangled each","In experiments, our method successfully disentangled the rationale of each move during the Go game.",3198
neural data. methods networks on convolutional from task learning allow based the Recent entire (CNN),Recent methods based on convolutional neural networks (CNN) allow learning the entire task from data.,3199
"frameworks. strengths we combines of the RayNet, paper, propose this In both which","In this paper, we propose RayNet, which combines the strengths of both frameworks.",3200
using RayNet train risk end-to-end We empirical minimization.,We train RayNet end-to-end using empirical risk minimization.,3201
"second the also a includes coherent of is which Besides, jammers, scenario, presence addressed.","Besides, a second scenario, which also includes the presence of coherent jammers, is addressed.",3202
"explore and In this between trajectories people their orientations. we correlation work, the head","In this work, we explore the correlation between people trajectories and their head orientations.",3203
"particularly people slowly, MX-LSTM i.e. effective when is move","MX-LSTM is particularly effective when people move slowly, i.e.",3204
all the scenario models. challenging for most other,the most challenging scenario for all other models.,3205
accurate predictions longer for The proposed approach horizon. also on a allows time,The proposed approach also allows for accurate predictions on a longer time horizon.,3206
"image emerging been deblurring. for a have recently, powerful neural networks Most as tool deep","Most recently, deep neural networks have been emerging as a powerful tool for image deblurring.",3207
"this that paper, tasks. architecture for encoder-decoder better image deblurring we In gives prove results","In this paper, we prove that encoder-decoder architecture gives better results for image deblurring tasks.",3208
on Recent based mostly Networks Detection (SOD) Salient (CNNs). are Convolutional Neural systems Object,Recent Salient Object Detection (SOD) systems are mostly based on Convolutional Neural Networks (CNNs).,3209
also background task the information the in is of ignored SOD. significant But,But the ignored background information is also significant in the task of SOD.,3210
object training solve by problem the for detection. introducing We to designed this data try,We try to solve this problem by introducing the training data designed for object detection.,3211
can slightly improve large-scale images the The performance of object of SOD.,The large-scale of object images can slightly improve the performance of SOD.,3212
Experiments immediately euthanasia. on cadavers Approach: were following performed in-vitro,Approach: Experiments were performed in-vitro on cadavers immediately following euthanasia.,3213
introduced Tukey ringing was A to window phenomena. Gibbs from remove,A Tukey window was introduced to remove ringing from Gibbs phenomena.,3214
object Scale is key challenges in detection. the of variation one,Scale variation is one of the key challenges in object detection.,3215
purposes. This for approach is beneficial triage,This approach is beneficial for triage purposes.,3216
is neuroanatomical-functional essential relationships. magnetic imaging resonance structural understanding on Whole (MRI) for segmentation brain,Whole brain segmentation on structural magnetic resonance imaging (MRI) is essential for understanding neuroanatomical-functional relationships.,3217
"segmentation. segmentation standard has method whole Traditionally, multi-atlas brain as for the been regarded","Traditionally, multi-atlas segmentation has been regarded as the standard method for whole brain segmentation.",3218
"at scans subject Moreover, from protocols. each two MRI different has least","Moreover, each subject has at least two scans from different MRI protocols.",3219
"the assess reproducibility. method's intra- and Herein, SLANT we inter-protocol","Herein, we assess the SLANT method's intra- and inter-protocol reproducibility.",3220
that SLANT intra- method achieved high reproducibility. protocol and inter- the The show results,The results show that the SLANT method achieved high intra- and inter- protocol reproducibility.,3221
of the We individual experimentally. thoroughly architecture components evaluate the and the contributions,We thoroughly evaluate the architecture and the contributions of the individual components experimentally.,3222
"horses prey such Furthermore, animals tend pain. hide as to their","Furthermore, prey animals such as horses tend to hide their pain.",3223
human comparisons state-of-the-art We versions generalized of additionally perform pain recognition with methods. baseline,We additionally perform baseline comparisons with generalized versions of state-of-the-art human pain recognition methods.,3224
of The this three-fold. purpose is work,The purpose of this work is three-fold.,3225
contributions inference new for context in musculo-skeletal disorder force management. These perspectives offer of the,These contributions offer new perspectives for force inference in the context of musculo-skeletal disorder management.,3226
"videos The direct portrait society, privacy. of law, threats recent and proliferation poses fake on","The recent proliferation of fake portrait videos poses direct threats on society, law, and privacy.",3227
"deep other a words, detector. fake introduce In we","In other words, we introduce a deep fake detector.",3228
position paper saying background discusses artificial the in the This (AI). of intelligence,This position paper discusses the saying in the background of artificial intelligence (AI).,3229
examining objectiveness proposed the rules the is of methodology meta evaluations. for A of,A methodology of meta rules is proposed for examining the objectiveness of the evaluations.,3230
presented. of also The possible saying problems are the,The possible problems of the saying are also presented.,3231
"by affected multipath biases. quality Nevertheless, and positioning clock is","Nevertheless, positioning quality is affected by multipath and clock biases.",3232
or bistatic acquired repeat-pass pairs mode. image in,image pairs acquired in repeat-pass or bistatic mode.,3233
"algorithm, Eclat, mining categorical itemset (FIM) accommodate to a frequent We modify variables.","We modify Eclat, a frequent itemset mining (FIM) algorithm, to accommodate categorical variables.",3234
hold out proposed method forecasting baselines. show samples On the we that outperforms these,On hold out samples we show that the proposed forecasting method outperforms these baselines.,3235
of We limitations by benefits conclude proposed approach. analyzing the and,We conclude by analyzing benefits and limitations of the proposed approach.,3236
spatial controllability lack translation. during also Existing of methods,Existing methods also lack of spatial controllability during translation.,3237
"to images. impose segmentor is on network particular, a In the designed information generated semantic","In particular, a segmentor network is designed to impose semantic information on the generated images.",3238
"However, supervised time-consuming and such collecting data labor-intensive. is","However, collecting such supervised data is time-consuming and labor-intensive.",3239
portfolio propose teach to manage from to a scratch. agent planning-based an We method,We propose a planning-based method to teach an agent to manage portfolio from scratch.,3240
with techniques Our learning deep techniques search combines reinforcement AlphaGo. approach like,Our approach combines deep reinforcement learning techniques with search techniques like AlphaGo.,3241
optimize utility combinatorial function The can search. algorithm by also non-differentiable suggested,The suggested algorithm can also optimize non-differentiable utility function by combinatorial search.,3242
then trading technique our used is This system. in,This technique is then used in our trading system.,3243
explore We the pose contextual importance of human information in estimation. spatial,We explore the importance of spatial contextual information in human pose estimation.,3244
to informative Cascade maps signals. from previous accumulates prediction fusion stages extract prediction,Cascade prediction fusion accumulates prediction maps from previous stages to extract informative signals.,3245
also prediction guide following at a to prior maps as The stages. resulting function,The resulting maps also function as a prior to guide prediction at following stages.,3246
joints Direct is captured. passing spatial is enabled different and relation between message,Direct message passing between different joints is enabled and spatial relation is captured.,3247
computational two limited These very require modules complexity.,These two modules require very limited computational complexity.,3248
results method and Experimental LSP consistently that on MPII benchmark. outperforms methods previous our demonstrate,Experimental results demonstrate that our method consistently outperforms previous methods on MPII and LSP benchmark.,3249
"the strategy an we this results alternative In interpret of paper, introduce to a CNN.","In this paper, we introduce an alternative strategy to interpret the results of a CNN.",3250
to networks then detection. adversarial We our interpretable BoW example for propose exploit,We then propose to exploit our interpretable BoW networks for adversarial example detection.,3251
across Convolutional achieved significant computer vision tasks. Networks Neural multiple have success,Convolutional Neural Networks have achieved significant success across multiple computer vision tasks.,3252
and vision industry other tasks It the in research to community. promote is expected,It is expected to promote other vision tasks in the research and industry community.,3253
of Web and virtually data. provide unlimited a platforms Social amount Media this multimodal,Web and Social Media platforms provide a virtually unlimited amount of this multimodal data.,3254
CP-PLL derived. second-order of non-linear paper rigorously In this model is a,In this paper a non-linear second-order model of CP-PLL is rigorously derived.,3255
of of model obtained the obviates second-order previously known The shortcomings CP-PLL. models,The obtained model obviates the shortcomings of previously known second-order models of CP-PLL.,3256
for is the CP-PLL. Pull-in second-order estimated time obtained,Pull-in time is estimated for the obtained second-order CP-PLL.,3257
potential Emerging biomedical significant (DL)-based have imaging. deep-learning to revolutionize techniques,Emerging deep-learning (DL)-based techniques have significant potential to revolutionize biomedical imaging.,3258
"temporally phenomena. identify Furthermore, and we show rare low-certainty spatially regions can that biological","Furthermore, we show that low-certainty regions can identify spatially and temporally rare biological phenomena.",3259
the attempts in models. have learning various explaining There been general numerous cognitive behaviours by,There have been numerous attempts in explaining the general learning behaviours by various cognitive models.,3260
show is with experiments optimization Our other that approaches. method competitive our based,Our experiments show that our method is competitive with other optimization based approaches.,3261
"antennas. modulation this reconfigurable the of on study paper, spatial based we In performance","In this paper, we study the performance of spatial modulation based on reconfigurable antennas.",3262
"increasingly visual recovering become completion has tensor in Recently, data. low-rank attractive incomplete","Recently, low-rank tensor completion has become increasingly attractive in recovering incomplete visual data.",3263
case. is matrix nuclear generalized tensor The case to the the truncated norm from,The truncated nuclear norm is generalized from the matrix case to the tensor case.,3264
branches is learning. and learning in one the machine new Deep important of,Deep learning is one of the new and important branches in machine learning.,3265
framework Learning a algorithms. Deep several contains is that important,Deep Learning is a framework that contains several important algorithms.,3266
"results as networks the different above, are we and from said models very However, different.","However, as we said above, the results from different networks and models are very different.",3267
"product accuracy, a build After a getting with we model high model.","After getting a model with high accuracy, we build a product model.",3268
the highest MNIST also we and achieved level. LeNet sparsity architecture speedup For and,For MNIST and LeNet architecture we also achieved the highest sparsity and speedup level.,3269
varying algorithm a We present of dominance characters virtual generating with data-driven gaits for traits.,We present a data-driven algorithm for generating gaits of virtual characters with varying dominance traits.,3270
relationships our present an characters. application virtual between that approach We simulates also of interpersonal,We also present an application of our approach that simulates interpersonal relationships between virtual characters.,3271
"data models. categorical panel or (quasi-)experimental data regressors,","categorical regressors, (quasi-)experimental data or panel data models.",3272
estimation. and risk-optimal plug-in a We parameters penalty propose approach derive for,We derive risk-optimal penalty parameters and propose a plug-in approach for estimation.,3273
asymptotic provide of schemes. different penalization distributions We under shrinkage the the estimators,We provide the asymptotic distributions of the shrinkage estimators under different penalization schemes.,3274
finite reveal standard samples. simulations over Monte robust improvements Carlo methods in,Monte Carlo simulations reveal robust improvements over standard methods in finite samples.,3275
"situations. temporal used many are In model real to temporal network different world planning, formalisms","In temporal planning, many different temporal network formalisms are used to model real world situations.",3276
to difficulty planning poses many reinforcement major a algorithms. learning Long-term,Long-term planning poses a major difficulty to many reinforcement learning algorithms.,3277
pronounced more environments. visual becomes in dynamic This even problem,This problem becomes even more pronounced in dynamic visual environments.,3278
annotated voxels. used maps to the are train classifier a Those with supervised,Those maps are used to train a supervised classifier with the annotated voxels.,3279
our setup have compared et Kiros al. We with,We have compared our setup with Kiros et al.,3280
problem. Many low-rank approximation previous formulate as issue a matrix this approaches,Many previous approaches formulate this issue as a low-rank matrix approximation problem.,3281
(TNNR) in method truncated real-world The applicable nuclear norm scenarios. regularization is,The truncated nuclear norm regularization (TNNR) method is applicable in real-world scenarios.,3282
the the is The values than to TNNR. singular of number robust DW-TNNR more truncated,The DW-TNNR is more robust to the number of truncated singular values than the TNNR.,3283
"In mechanism attention supervision. addition, to introduce the provide also additional we","In addition, we also introduce the attention mechanism to provide additional supervision.",3284
a but future fantasy Predicting work. practicality is the,Predicting the future is a fantasy but practicality work.,3285
inter-frame that synthetic is The predictions explicitly enforce learns future-frame to guided difference. by DGGAN,The DGGAN learns to explicitly enforce future-frame predictions that is guided by synthetic inter-frame difference.,3286
"a to paths dual meta Given information. first generate DGGAN uses of frames, sequence","Given a sequence of frames, DGGAN first uses dual paths to generate meta information.",3287
moving motion a from for We stereo learning independently system present objects videos. of,We present a system for learning motion of independently moving objects from stereo videos.,3288
constraints Leveraging renders or via direct flow supervision maps depth geometric unnecessary.,Leveraging geometric constraints renders direct supervision via depth or flow maps unnecessary.,3289
"fuse depth-guided Our a per-pixel predictions. network and mask, predicts to pixel-based used furthermore","Our network furthermore predicts a per-pixel mask, used to fuse depth-guided and pixel-based predictions.",3290
"First, spatial map. both explore map and we eye-fixation saliency in correlations","First, we explore spatial correlations in both saliency map and eye-fixation map.",3291
"a define Then, we that reliable saliency spatiotemporal learn correlations map.","Then, we learn spatiotemporal correlations that define a reliable saliency map.",3292
"In experiments, of we definitions signals. video and our in locality spatiotemporal neighborhoods multiple explore","In our experiments, we explore multiple definitions of locality and neighborhoods in spatiotemporal video signals.",3293
"algorithm proposed making for more natural generalization suitable is The most it to unsupervised, videos.","The proposed algorithm is unsupervised, making it more suitable for generalization to most natural videos.",3294
"and customization it flexible content. is to video Also, for efficient computationally specific","Also, it is computationally efficient and flexible for customization to specific video content.",3295
"fast, iris matching performance in is contingent segmentation. iris biometrics high-quality upon Subject","Subject matching performance in iris biometrics is contingent upon fast, high-quality iris segmentation.",3296
efficient required designs effective provision. stack across achieve to QoS Novel and protocol the are,Novel designs across the protocol stack are required to achieve effective and efficient QoS provision.,3297
performance effectively future highly resulting efficiency requirements applications. can and design stringent satisfy The of,The resulting design can effectively satisfy highly stringent performance and efficiency requirements of future applications.,3298
a in image segmentation of processing. is Interactive many image topic studies,Interactive image segmentation is a topic of many studies in image processing.,3299
segmentation a graph-based This for interactive with method paper presents stages. new two,This paper presents a new graph-based method for interactive segmentation with two stages.,3300
"its proposed perform simplicity, accuracy. can Despite method the high task with the","Despite its simplicity, the proposed method can perform the task with high accuracy.",3301
This forms. perception ownership assignment in of step the border is important an,This border ownership assignment is an important step in the perception of forms.,3302
modern Region detection the cornerstone techniques. of are anchors object,Region anchors are the cornerstone of modern object detection techniques.,3303
"this stage. paper, foundational this revisit we In","In this paper, we revisit this foundational stage.",3304
study more that and effectively Our be it can shows done much efficiently.,Our study shows that it can be done much more effectively and efficiently.,3305
performance. We use to study improve also the high-quality proposals of detection,We also study the use of high-quality proposals to improve detection performance.,3306
scheme proposal into seamlessly integrated be can detectors. and methods The anchoring,The anchoring scheme can be seamlessly integrated into proposal methods and detectors.,3307
over an set of consider signal vectors. a We of overcomplete decomposition the,We consider the decomposition of a signal over an overcomplete set of vectors.,3308
in strong prevalent is for infancy animal across emerges A novelty the preference kingdom. and,A strong preference for novelty emerges in infancy and is prevalent across the animal kingdom.,3309
dequantization. novel for architecture a propose first CNN color We,We first propose a novel CNN architecture for color dequantization.,3310
GIF of adapt temporal We then network frames. the for interpolation SuperSlomo,We then adapt the SuperSlomo network for temporal interpolation of GIF frames.,3311
"evaluation. and two namely introduce GIF-Faces and for We training large GIF-Moments, both datasets,","We introduce two large datasets, namely GIF-Faces and GIF-Moments, for both training and evaluation.",3312
"methods, expressions of derived. both probability are closed-form For error","For both methods, closed-form expressions of error probability are derived.",3313
"Monte-Carlo verified by simulations. analytical Finally, expressions are","Finally, analytical expressions are verified by Monte-Carlo simulations.",3314
"scar a morphological by operations means is segmentation coarse of fast obtained. Secondly,","Secondly, by means of morphological operations a fast coarse scar segmentation is obtained.",3315
"a the by segmentation Thirdly, refined CNNs. using strategy boundary-voxel an of reclassification ensemble is","Thirdly, the segmentation is refined by a boundary-voxel reclassification strategy using an ensemble of CNNs.",3316
"reproduce ranges. Moreover, intra- variability the method the able inter-observer was to and","Moreover, the method was able to reproduce the intra- and inter-observer variability ranges.",3317
transferred could is method be the clinical that to concluded scenarios. suitably It,It is concluded that the method could suitably be transferred to clinical scenarios.,3318
these improvements several algorithmic artifacts. made We remove to,We made several algorithmic improvements to remove these artifacts.,3319
right software community. by the and usable algorithm out-of-the-box pathology is Our computational,Our algorithm and software is usable right out-of-the-box by the computational pathology community.,3320
"Amongst immune collectins them, players important defense. the in are antiviral innate","Amongst them, collectins are important players in the antiviral innate immune defense.",3321
SP-D were for porcine effects (pSP-D). Similar observed,Similar effects were observed for porcine SP-D (pSP-D).,3322
thickness biomarker used volume commonly imaging such Its or is as a diseases. surrogate for,Its volume or thickness is commonly used as a surrogate imaging biomarker for such diseases.,3323
what face different of filters a Do know we the network represent?,Do we know what the different filters of a face network represent?,3324
use to transfer other without train this tasks filter learning? Can we information,Can we use this filter information to train other tasks without transfer learning?,3325
case the These estimators well in established the real are studied since seventies. and,These estimators are well established and studied in the real case since the seventies.,3326
interest. Their has drawn recent to extension the case complex,Their extension to the complex case has drawn recent interest.,3327
and vision of the SalSi human theory saliency the based system modeling (HVS). is on,SalSi is based on the saliency theory and modeling of the human vision system (HVS).,3328
itself a diverse pool curating informative challenge. a Yet sufficiently images of and is,Yet curating a pool of sufficiently diverse and informative images is itself a challenge.,3329
generation propose to active approach this issue. an image address We,We propose an active image generation approach to address this issue.,3330
"system to without central result, collaborative do modeling As any server. users a our enables","As a result, our system enables users to do collaborative modeling without any central server.",3331
of social make images machines in better beings. Discovering behavior the can relations interpret human,Discovering social relations in images can make machines better interpret the behavior of human beings.,3332
"is graphs, graph performed social convolutional reasoning networks. relation the by Based on","Based on the graphs, social relation reasoning is performed by graph convolutional networks.",3333
Extensive proposed on the framework. two the experiments datasets of public show effectiveness,Extensive experiments on two public datasets show the effectiveness of the proposed framework.,3334
fading learn intelligently is can the statistics because of DNN the This channels.,This is because DNN can intelligently learn the statistics of the fading channels.,3335
"is indexer, the crawler, the and PVSS modules: the constituted searcher. three of","PVSS is constituted of three modules: the crawler, the indexer, and the searcher.",3336
by the will searched in target process. a progressive be vehicle The database,The target vehicle will be searched in the database by a progressive process.,3337
We data finetune real pretraining after data. synthetic with with,We finetune with real data after pretraining with synthetic data.,3338
"a the features texture. document handwritten new describing present this In we some paper, as","In this paper, we present some new features describing the handwritten document as a texture.",3339
are on the Radon transform. features based These,These features are based on the Radon transform.,3340
and obtained coarse documents. for classification can of easily suit be values the All,All values can be obtained easily and suit for the coarse classification of documents.,3341
"existing perform allows network us The to approaches. part-level resulting manipulation, unattainable shape by","The resulting network allows us to perform part-level shape manipulation, unattainable by existing approaches.",3342
We contextual in objects for propose model detecting adversarial an moving images.,We propose an adversarial contextual model for detecting moving objects in images.,3343
"However, position shift problem, the multispectral i.e. suffers from real data","However, real multispectral data suffers from the position shift problem, i.e.",3344
state-of-the-art outperforms experimental result DSMHN significantly The that demonstrates methods.,The experimental result demonstrates that DSMHN significantly outperforms state-of-the-art methods.,3345
multiple parts. consist objects meaningful of semantically Hand-drawn usually,Hand-drawn objects usually consist of multiple semantically meaningful parts.,3346
"model into segments neural symbols In this stroke-level that we a propose network components. paper,","In this paper, we propose a neural network model that segments symbols into stroke-level components.",3347
the surpass Our of state art existing an segmentation small dataset. available methodologies on scores,Our segmentation scores surpass existing methodologies on an available small state of the art dataset.,3348
the the We release to community. dataset,We release the dataset to the community.,3349
"devices IoT this real time. approach, single-batch in to inferencing cooperate In conduct","In this approach, IoT devices cooperate to conduct single-batch inferencing in real time.",3350
the paper addresses interpolating This of visual textures. problem,This paper addresses the problem of interpolating visual textures.,3351
several robustness public on of datasets Experiments demonstrate the our and method. efficiency,Experiments on several public datasets demonstrate the efficiency and robustness of our method.,3352
terrestrial of and networks. differentiate aqueous structure the to these between,aqueous and terrestrial to differentiate between the structure of these networks.,3353
sources allows images. to GI one in color-biased estimate multiple or illumination,GI allows to estimate one or multiple illumination sources in color-biased images.,3354
"it shapes recognize and However, patterns. considerably to difficult of is distorted because various its","However, it is considerably difficult to recognize because of its various shapes and distorted patterns.",3355
and rectification network attention-based sequence of a recognition MORAN consists an network. The multi-object,The MORAN consists of a multi-object rectification network and an attention-based sequence recognition network.,3356
network for rectifying is multi-object contain The text. that images irregular designed rectification,The multi-object rectification network is designed for rectifying images that contain irregular text.,3357
the attention-based focuses and target outputs predictions. network sequentially on characters sequence recognition The,The attention-based sequence recognition network focuses on target characters and sequentially outputs the predictions.,3358
"both MORAN read text. irregular With regular mechanism, scene can the and the rectification","With the rectification mechanism, the MORAN can read both regular and irregular scene text.",3359
"which various achieves that on state-of-the-art show the MORAN benchmarks are performance. Extensive conducted, experiments","Extensive experiments on various benchmarks are conducted, which show that the MORAN achieves state-of-the-art performance.",3360
structured learning Machine real-world in data systems with applications. deal regularly,Machine learning systems regularly deal with structured data in real-world applications.,3361
"real-valued a vector of fixed, a pre-specified as size.","as a real-valued vector of a fixed, pre-specified size.",3362
on adhesion the models We cell-cell assumptions. continuum underlying several discuss based microscopic,We discuss several continuum cell-cell adhesion models based on the underlying microscopic assumptions.,3363
due cell-cell We finally to resulting repulsion. the behavior qualitative explore,We finally explore the resulting qualitative behavior due to cell-cell repulsion.,3364
can influence This highlights the how curvature the Errera tissue rule. model of idealized,This idealized model highlights how the curvature of the tissue can influence Errera rule.,3365
close salt has petroleum reservoirs The identification of exploration of domes. relationship a the with,The exploration of petroleum reservoirs has a close relationship with the identification of salt domes.,3366
two-stage trade-off. detectors detectors the single-shot of surged accuracy-vs-speed have Recently in ahead,Recently two-stage detectors have surged ahead of single-shot detectors in the accuracy-vs-speed trade-off.,3367
are Nevertheless detectors vision popular single-shot in applications. immensely embedded,Nevertheless single-shot detectors are immensely popular in embedded vision applications.,3368
techniques. brings current This same to the single-shot up paper two-stage as detectors level,This paper brings single-shot detectors up to the same level as current two-stage techniques.,3369
the call We resulting RetinaMask. network augmented,We call the resulting augmented network RetinaMask.,3370
to attacks.  vulnerable recognition Prevailing are systems spoof  fingerprint,  Prevailing fingerprint recognition systems are vulnerable to spoof attacks.,3371
"detection classification To approach one-class problem. a alleviate shortcoming, we this spoof as","To alleviate this shortcoming, we approach spoof detection as a one-class classification problem.",3372
"relatively strict require applied procedures are control, use. for that conditions process In statistical their","In statistical process control, procedures are applied that require relatively strict conditions for their use.",3373
"detection however, usually the the causes. increase charts, delay assignable control of in Robust","Robust control charts, however, usually increase the delay in the detection of assignable causes.",3374
robust to solve approach propose fast and this We a problem.,We propose a fast and robust approach to solve this problem.,3375
"geometric and to cues Moreover, propose combine matching. we cross-view appearance for","Moreover, we propose to combine geometric and appearance cues for cross-view matching.",3376
Photorealism easily concept be is complex a that cannot formulated mathematically.,Photorealism is a complex concept that cannot easily be formulated mathematically.,3377
the used measure aesthetics To image and elaborated. results is for a improve further,To further improve the results a measure for image aesthetics is used and elaborated.,3378
autonomous state for Accurate estimation a robots. fundamental is problem,Accurate state estimation is a fundamental problem for autonomous robots.,3379
"Within optimization, aligned coordinate. local estimations are a global into graph the","Within the graph optimization, local estimations are aligned into a global coordinate.",3380
evaluate real-world public system of with datasets the We experiments. performance our on and,We evaluate the performance of our system on public datasets and with real-world experiments.,3381
algorithms. state-of-the-art other against compared are Results,Results are compared against other state-of-the-art algorithms.,3382
"challenging. an is Experimentally, however, estimation PRCs experimental of accurate","Experimentally, however, an accurate experimental estimation of PRCs is challenging.",3383
representation the lower top-level image the of The combined and final of embeddings. embedding is,The final embedding of the image is combined representation of the lower and top-level embeddings.,3384
"our Also, compared current we art state methods. work with of the","Also, we compared our work with current state of the art methods.",3385
"ergonomic of them, in constraints. front to due","in front of them, due to ergonomic constraints.",3386
"usable for classification is strategy automated Thus, diagnosis. the proposed keratocyst patch","Thus, the proposed patch classification strategy is usable for automated keratocyst diagnosis.",3387
"However, make performed suitable to further for optimization use. must practical be it","However, further optimization must be performed to make it suitable for practical use.",3388
low-level and detectors build descriptors tools on. are that tasks key many vision Feature higher-level,Feature detectors and descriptors are key low-level vision tools that many higher-level tasks build on.,3389
high regime. results the information-guided The NUSM/NUBM in the is optimal reveal comparison that SNR,The comparison results reveal that the information-guided NUSM/NUBM is optimal in the high SNR regime.,3390
to analysis. are results Numerical presented validate our,Numerical results are presented to validate our analysis.,3391
"more equipped sensors on autonomous are more and increase and robustness ability. Nowadays, robots to","Nowadays, more and more sensors are equipped on robots to increase robustness and autonomous ability.",3392
employed of sensor Few with them multiple be choices. can,Few of them can be employed with multiple sensor choices.,3393
Every as our sensor a general treated factor framework. in is,Every sensor is treated as a general factor in our framework.,3394
common Factors problem. state are share build which optimization summed together the variables to,Factors which share common state variables are summed together to build the optimization problem.,3395
algorithms. state-of-the-art are against other compared Results,Results are compared against other state-of-the-art algorithms.,3396
"for In series this downlink novel we a paper, propose of communication. multi-cell coordination schemes","In this paper, we propose a series of novel coordination schemes for multi-cell downlink communication.",3397
"exchange is sharing data coordination CSI involves the need and only for the eliminated. Accordingly,","Accordingly, the coordination only involves CSI exchange and the need for sharing data is eliminated.",3398
"imperfect all CSI proposed the For considered. schemes, is","For all the proposed schemes, imperfect CSI is considered.",3399
terms deterministic transmission the of total power optimizations. minimize We and probabilistic in,We minimize the total transmission power in terms of probabilistic and deterministic optimizations.,3400
to novel it generative images. when synthesising have models great Deep shown comes promise,Deep generative models have shown great promise when it comes to synthesising novel images.,3401
"to metrics procedure training adapted The are the of task network, and fine-grained error sketching.","The network, error metrics and training procedure are adapted to the task of fine-grained sketching.",3402
publicly dataset We will available. our make,We will make our dataset publicly available.,3403
"this a we overcome cure GreedyNMS. to Pairwise-NMS To propose drawback,","To overcome this drawback, we propose a Pairwise-NMS to cure GreedyNMS.",3404
transformation. image-to-image focuses suitable of the the objective over for function research Most,Most of the research focuses over the suitable objective function for image-to-image transformation.,3405
"Adversarial this Networks a new we image-to-image transformation. In Cyclic-Synthesized for paper, Generative propose (CSGAN)","In this paper, we propose a new Cyclic-Synthesized Generative Adversarial Networks (CSGAN) for image-to-image transformation.",3406
Content-based detection role important in preventing adult pornography. video an plays,Content-based adult video detection plays an important role in preventing pornography.,3407
"usually However, multi-modality existing seldom semantics and on focus representation. modality single rely methods on","However, existing methods usually rely on single modality and seldom focus on multi-modality semantics representation.",3408
"patterns first, in and are visual-frames. salient and analyzed respective-ly At periodic regions audio-frames","At first, periodic patterns and salient regions are respective-ly analyzed in audio-frames and visual-frames.",3409
"periodicity visual multi-modal by the with combining audio is co-occurrence saliency. Next, described semantics","Next, the multi-modal co-occurrence semantics is described by combining audio periodicity with visual saliency.",3410
"step. performance of the Moreover, step evaluated by is approach our","Moreover, the performance of our approach is evaluated step by step.",3411
outper-forms our results Experimental that state-of-the-art approach some obviously show methods.,Experimental results show that our approach obviously outper-forms some state-of-the-art methods.,3412
"studied well previous works. in However, this has not been problem","However, this problem has not been well studied in previous works.",3413
richer results than finer able to is model details annotations. has produce which segmentation Our,Our model is able to produce finer segmentation results which has richer details than annotations.,3414
single best not liver a algorithm and found that We for performed tumors.,We found that not a single algorithm performed best for liver and tumors.,3415
accuracy. DCT the detection showed highest method experimental that has the vehicle The results,The experimental results showed that the DCT method has the highest vehicle detection accuracy.,3416
hard these is past. because includes the signs the Capturing context,Capturing these signs is hard because the context includes the past.,3417
color to and algorithm image denoising. MBWPNM the adopt then multispectral We,We then adopt the MBWPNM algorithm to color and multispectral image denoising.,3418
make There them of are that attractive. several strengths methods imaging,There are several strengths of imaging methods that make them attractive.,3419
"low-cost ubiquitous cameras solutions for often Furthermore, are sensing. and","Furthermore, cameras are ubiquitous and often low-cost solutions for sensing.",3420
We the to that physiological of will sensing this hope non-contact advancement toolbox contribute methods.,We hope that this toolbox will contribute to the advancement of non-contact physiological sensing methods.,3421
systems. interaction hand gesture-based segmentation role an Hand detection play indispensable fingertip in human-machine and,Hand segmentation and fingertip detection play an indispensable role in hand gesture-based human-machine interaction systems.,3422
image a foreground is in editing. Replacing the objects challenging and background simultaneously adjusting task,Replacing the background and simultaneously adjusting foreground objects is a challenging task in image editing.,3423
"a network, called fish-like we goal, design FishNet. this Towards","Towards this goal, we design a fish-like network, called FishNet.",3424
"the information for task. the final and of is preserved resolutions In refined FishNet, all","In FishNet, the information of all resolutions is preserved and refined for the final task.",3425
this better problem. can design handle Our,Our design can better handle this problem.,3426
demonstrate to experiments of Extensive the been have the remarkable performance FishNet. conducted,Extensive experiments have been conducted to demonstrate the remarkable performance of the FishNet.,3427
fast retrieval. image reduction for coded dimension The features are hash binary after,The features after dimension reduction are binary hash coded for fast image retrieval.,3428
"concepts complex internal creating for recent the we present of auditory models environment. cognitive First,","First, we present recent cognitive concepts for creating internal models of the complex auditory environment.",3429
"lymphatic transport Cranial of fluids, immune CNS (LVs) and macromolecules in responses. are involved vessels","Cranial lymphatic vessels (LVs) are involved in transport of fluids, macromolecules and CNS immune responses.",3430
extensively and remodel LVs injury. are Spinal spinal after cord VEGF-C-dependent,Spinal LVs are VEGF-C-dependent and remodel extensively after spinal cord injury.,3431
in and order then to is the separate subtraction foreground. applied background the Background,Background subtraction is then applied in order to separate the background and the foreground.,3432
"of in camera, are investigated environments. foreground challenges objects Thus, terms and","Thus, challenges are investigated in terms of camera, foreground objects and environments.",3433
point correspondences views. using three these relationship among further We geometrical refine,We further refine these point correspondences using geometrical relationship among three views.,3434
on slides. based tissue of visual examination cancer Histopathological diagnosis is stained,Histopathological cancer diagnosis is based on visual examination of stained tissue slides.,3435
routinely is and Hematoxylin eosin worldwide. (H\&E) employed stain standard a,Hematoxylin and eosin (H\&E) is a standard stain routinely employed worldwide.,3436
the IHC slide. can proteins staining These be by target of limitations overcome of tissue,These limitations can be overcome by IHC staining of target proteins of the tissue slide.,3437
"are challenging, is there supervision. no available example translations as This especially when","This is challenging, especially when there are no example translations available as supervision.",3438
"and Prior each mix a train generated with network comparison, data. of works, in real","Prior works, in comparison, train each network with a mix of real and generated data.",3439
"into discuss algorithms, then them broad three grouping reconstruction generations. We","We then discuss reconstruction algorithms, grouping them into three broad generations.",3440
"video frames computing is time-consuming. However, very from flow","However, computing flow from video frames is very time-consuming.",3441
performance by good and layouts cross-validation. Our general shows on results approach qualitative,Our approach shows good performance on general layouts by qualitative results and cross-validation.,3442
mismatch. Most optimization attempt output-target minimize an methods to that problems such solve,Most such methods attempt to solve optimization problems that minimize an output-target mismatch.,3443
"Often, and given. are naturally not handcrafted rewards those","Often, those rewards are handcrafted and not naturally given.",3444
"we the related at factors have risk. key to students Then, found strongly","Then, we have found key factors strongly related to the students at risk.",3445
"drop out. order it in is not Therefore, the to lectures attend crucial to","Therefore, it is crucial to attend the lectures in order not to drop out.",3446
show has a that Experiment remarkable effect. results theory approximation the asymptotic,Experiment results show that the asymptotic theory has a remarkable approximation effect.,3447
"are using validated Monte simulation. the Finally, expressions Carlo mathematical the","Finally, the mathematical expressions are validated using the Monte Carlo simulation.",3448
the has The to method applied been matching stereo problem. proposed,The proposed method has been applied to the stereo matching problem.,3449
unorganized images is reconstruction computer task many from in important vision RGB applications. Scene an,Scene reconstruction from unorganized RGB images is an important task in many computer vision applications.,3450
for the practically neural often useful sampling of data. Numerical techniques decoding real are,Numerical sampling techniques are often practically useful for the decoding of real neural data.,3451
over parameters. characterized We by neural neuron sensory consider distribution a populations,We consider sensory neural populations characterized by a distribution over neuron parameters.,3452
the comparison filtering with approximation. quality of Numerical particle the demonstrates,Numerical comparison with particle filtering demonstrates the quality of the approximation.,3453
detection news role of in an important video analysis. videos news plays Format,Format detection of news videos plays an important role in news video analysis.,3454
the Probabilistic progressive used edges. band Hough detection for transform of has been,Probabilistic progressive Hough transform has been used for the detection of band edges.,3455
"generated text natural images, bands detected are and The (non-text) as classified bands. computer graphics","The detected bands are classified as natural images, computer generated graphics (non-text) and text bands.",3456
per-point learn define features. a loss to Our self-supervision function approach rich to relevant leverages,Our approach leverages self-supervision to define a relevant loss function to learn rich per-point features.,3457
information We point's geometric into of local use to latent incorporate representation. each structures point-clouds,We use local structures of point-clouds to incorporate geometric information into each point's latent representation.,3458
through the applications. point-wise three our meaningful We to capture demonstrate ability method of features,We demonstrate the ability of our method to capture meaningful point-wise features through three applications.,3459
"embedding space, By perform clouds. learned we unsupervised the clustering point part-segmentation on","By clustering the learned embedding space, we perform unsupervised part-segmentation on point clouds.",3460
semantic we latent in point-analogies. derive By distance the euclidean calculating space,By calculating euclidean distance in the latent space we derive semantic point-analogies.,3461
effectively layers in networks. neural hidden captured of are,are effectively captured in hidden layers of neural networks.,3462
The image main a deep architecture Encoder-Decoder learning segmentation. is biomedical for stream model,The Encoder-Decoder architecture is a main stream deep learning model for biomedical image segmentation.,3463
"such are still decoders under-explored architectures. in However,","However, decoders are still under-explored in such architectures.",3464
"active acquire As network data, is in network nature. learns the our to","As our network learns to acquire data, the network is active in nature.",3465
Carlo signals supervision two created search are Monte through The tree (MCTS).,The two supervision signals are created through Monte Carlo tree search (MCTS).,3466
Digital analysis remote image or enables algorithms. pathology and access consults powerful,Digital pathology enables remote access or consults and powerful image analysis algorithms.,3467
"(OOF). process digitization However, can such artifacts as create the slide out-of-focus","However, the slide digitization process can create artifacts such as out-of-focus (OOF).",3468
"potentially is only detected OOF workflow upon review, rescanning careful and delays. causing often","OOF is often only detected upon careful review, potentially causing rescanning and workflow delays.",3469
multiple of We approach with computational the cost also our evaluated measurements.,We also evaluated the computational cost of our approach with multiple measurements.,3470
effectively large monitoring for in be scale used screening. methods temperature patient Contactless could,Contactless temperature monitoring methods could be used effectively in large scale for patient screening.,3471
Near-infrared convenient thermography to be for has and accurate temperature profiling. proven,Near-infrared thermography has proven to be convenient and accurate for temperature profiling.,3472
cancer leading death Lung cause is of worldwide. Objective: the cancer-related,Objective: Lung cancer is the leading cause of cancer-related death worldwide.,3473
"using sequencers radiomic interpretation challenge. a such predictions However, the of made remains","However, the interpretation of predictions made using such radiomic sequencers remains a challenge.",3474
on and correspondences. based state-of-the-art transformation estimation This synthetic proposals paper both eleven evaluates descriptor,This paper evaluates eleven state-of-the-art transformation estimation proposals on both descriptor based and synthetic correspondences.,3475
"these evaluate addition, In also methods. the we efficiencies of","In addition, we also evaluate the efficiencies of these methods.",3476
"estimation and merits, application of these the Finally, methods demerits tested transformation guidance summarized. are","Finally, the merits, demerits and application guidance of these tested transformation estimation methods are summarized.",3477
It is parameters each distinct segment. the to are that assume common within,It is common to assume that the parameters are distinct within each segment.,3478
"balanced clustering paper, consumption proposed. In (BECC) is a energy algorithm this","In this paper, a balanced energy consumption clustering algorithm (BECC) is proposed.",3479
scheme sensor new energy cluster-based algorithm heterogeneous a is designed networks. wireless This for,This new scheme is a cluster-based algorithm designed for heterogeneous energy wireless sensor networks.,3480
animation obtainable easily prior method uses Our data.,Our method uses easily obtainable prior animation data.,3481
can monocular a input. only our convincing We method video that show generate using animations,We show that our method can generate convincing animations using only a monocular video input.,3482
two-area of demonstrated Kundur is system. in system the proposed The effectiveness the,The effectiveness of the proposed system is demonstrated in the Kundur two-area system.,3483
"the are errors applied. measures as latter such mean objective square comparisons, For","For the latter comparisons, objective measures such as mean square errors are applied.",3484
estimation. optical quality mostly visual of on frames flow interpolated based,visual quality of interpolated frames mostly based on optical flow estimation.,3485
"of article introduces This suite programs regularized for Stata. a lassopack, in regression","This article introduces lassopack, a suite of programs for regularized regression in Stata.",3486
"and ridge elastic post-estimation square-root lassopack adaptive net, lasso, lasso regression, implements lasso, OLS.","lassopack implements lasso, square-root lasso, elastic net, ridge regression, adaptive lasso and post-estimation OLS.",3487
each theoretical approach. the framework practical considerations discuss for We and,We discuss the theoretical framework and practical considerations for each approach.,3488
of to penalization Carlo performance Monte present results compare the We also the approaches.,We also present Monte Carlo results to compare the performance of the penalization approaches.,3489
"during under-utilization the particularly testing, classification modalities, This of hinder can accuracy method. of the","This under-utilization of modalities, particularly during testing, can hinder the classification accuracy of the method.",3490
radio. still technology developing amateur still APRS among exists is and,APRS technology is still exists and still developing among amateur radio.,3491
"in APRS. Indonesia utilize still use there not people and many However, are who","However, in Indonesia there are still not many people who use and utilize APRS.",3492
for a non-matching The result in coalescent algorithm enumerating histories case. simplified the provides caterpillar,The result provides a simplified algorithm for enumerating coalescent histories in the non-matching caterpillar case.,3493
"These trade-offs efficiency fidelity, compression capabilities. and across provide","These provide trade-offs across fidelity, efficiency and compression capabilities.",3494
"the all be can subproblems solved analytically. algorithm, In the","In the algorithm, all the subproblems can be solved analytically.",3495
"subproblem in can be except subproblems performed Moreover, all parallel. one","Moreover, all subproblems except one subproblem can be performed in parallel.",3496
examples variants. consensus-ADMM and algorithm show proposed its Numerical effectiveness of the the,Numerical examples show the effectiveness of the proposed consensus-ADMM algorithm and its variants.,3497
conditions presented. are disease the for of and permanence The extinction the,The conditions for the extinction and permanence of the disease are presented.,3498
the algorithm. are offered and empirical results Both support to theoretical,Both theoretical and empirical results are offered to support the algorithm.,3499
achievable blocklength adopted. is finite rate under regime data The,The achievable data rate under finite blocklength regime is adopted.,3500
iterative algorithm problem. perturbation-based proposed is solve A this to novel,A novel perturbation-based iterative algorithm is proposed to solve this problem.,3501
the to requirements. design parameters We antenna optimal evaluate,We evaluate the optimal antenna parameters to design requirements.,3502
us estimate allows simultaneously. This the to deformations geometric correspondence and,This allows us to estimate the correspondence and geometric deformations simultaneously.,3503
from attention both Neural industrial transfer field. drawn has academic considerable and style,Neural style transfer has drawn considerable attention from both academic and industrial field.,3504
"transfer we limitations paper, style tackle by an developing this In attention-aware multi-stroke these model.","In this paper, we tackle these limitations by developing an attention-aware multi-stroke style transfer model.",3505
system A exponential analysis utilized sample obtain results. Lyapunov extinction the to for is,A sample Lyapunov exponential analysis for the system is utilized to obtain extinction results.,3506
develop procedure free on model a We a predicate. inference likelihood conditioning for a probabilistic,We develop a likelihood free inference procedure for conditioning a probabilistic model on a predicate.,3507
predicate function a a question is domain. expresses a which yes/no A valued Boolean about,A predicate is a Boolean valued function which expresses a yes/no question about a domain.,3508
thus method termed is MUSIC. one-bit as resulting The,The resulting method is thus termed as one-bit MUSIC.,3509
"However, for. the unaccounted still speckle leaves this","However, this still leaves the speckle unaccounted for.",3510
"the of to However, micro-expressions. changes spatiotemporal subtle challenging recognition still micro-expression is the due","However, the micro-expression recognition is still challenging due to the subtle spatiotemporal changes of micro-expressions.",3511
manual optimized and by an end-to-end manner It is design. feature obviates,It is optimized by an end-to-end manner and obviates manual feature design.,3512
train to DDPG use algorithm an expert We agent. original,We use DDPG algorithm to train an original expert agent.,3513
"detection splicing. plays image identifying an image in role such tampering, important Resampling as","Resampling detection plays an important role in identifying image tampering, such as image splicing.",3514
"the study, compare the versions a of smoother. In we different of simulation performance","In a simulation study, we compare the performance of different versions of the smoother.",3515
"hardware relays, ideal Many research assuming etc.) (source, attempts previous","Many previous research attempts assuming ideal hardware (source, relays, etc.)",3516
"valid fact, this for assumption low-rate still is systems. In","In fact, this assumption is still valid for low-rate systems.",3517
"presented. and simulation Finally, numerical results are","Finally, numerical and simulation results are presented.",3518
Each plug-and-play a iteration inversion involves denoising model by the step. followed the forward of,Each plug-and-play iteration involves the inversion of the forward model followed by a denoising step.,3519
"into result. the incorporate optimization framework a can hard Moreover, we as constraints easily","Moreover, we can easily incorporate hard constraints into the optimization framework as a result.",3520
and proposals for imaging. demonstrate We single-photon super-resolution our the effectiveness of,We demonstrate the effectiveness of our proposals for super-resolution and single-photon imaging.,3521
in image underwater proposed have algorithms last enhancement years. been few the Numerous,Numerous underwater image enhancement algorithms have been proposed in the last few years.,3522
"few selected datasets using real-world evaluated are However, mainly synthetic or images. algorithms these either","However, these algorithms are mainly evaluated using either synthetic datasets or few selected real-world images.",3523
statistics of performance the the Pareto modified algorithm. the of The data good shows HS,The statistics of the Pareto data shows the good performance of the modified HS algorithm.,3524
Hardware communication systems. main restrictions among are upcoming wireless for performance non-idealities the,Hardware non-idealities are among the main performance restrictions for upcoming wireless communication systems.,3525
parameters optimize iterative propose signals. transmit the We two to improper algorithms of the,We propose two iterative algorithms to optimize the parameters of the improper transmit signals.,3526
HWD. show two-user the We IC performance improve the IGS can with that of additive,We show that IGS can improve the performance of the two-user IC with additive HWD.,3527
image compression by is Ultrasound task. key a speckle-based challenging information preserving,Ultrasound image compression by preserving speckle-based key information is a challenging task.,3528
two building generalizability build We using variants demonstrate blocks of the to compressor. the,We demonstrate generalizability of the building blocks using two variants to build the compressor.,3529
We to as speckle compression work a standards. image envision this serve roadmap for frame,We envision this frame work to serve as a roadmap for speckle image compression standards.,3530
"video-based the problem this pedestrian In tackle approach. we first of recognition attribute paper, by","In this paper, we first tackle the problem of pedestrian attribute recognition by video-based approach.",3531
Pavement critical crack task is a insuring safety. for detection road,Pavement crack detection is a critical task for insuring road safety.,3532
"crack an detection boost this required to is road method automatic progress. Therefore,","Therefore, an automatic road crack detection method is required to boost this progress.",3533
images. Our of MS effectiveness the show significantly the results lesion of synthetic usage,Our results significantly show the effectiveness of the usage of synthetic MS lesion images.,3534
the counting applications transportation numerous Classifying traffic engineering has road in and in domain. vehicles,Classifying and counting vehicles in road traffic has numerous applications in the transportation engineering domain.,3535
"the (two-wheelers, cars, However, three-wheelers, of buses, variety vehicles etc.) trucks wide","However, the wide variety of vehicles (two-wheelers, three-wheelers, cars, buses, trucks etc.)",3536
exams. yield different techniques Digitization for images visual in radiological biomedical patterns,Digitization techniques for biomedical images yield different visual patterns in radiological exams.,3537
"different focus this datasets. paper, improving across of we In kinds on ability generalization the","In this paper, we focus on improving the generalization ability across different kinds of datasets.",3538
on silhouettes. We the and truly rasterizer non-parametric based first present differentiable,We present the first non-parametric and truly differentiable rasterizer based on silhouettes.,3539
results. challenging MOT is state-of-the-art the and benchmarks on method the The achieves evaluated,The method is evaluated on the challenging MOT benchmarks and achieves the state-of-the-art results.,3540
"solutions. numerical proposed examples, that designs we show, robust through outperform non-robust Furthermore, significantly the","Furthermore, we show, through numerical examples, that the proposed robust designs significantly outperform non-robust solutions.",3541
"method to lost Moreover, values. us the pixel proposed restore enables","Moreover, the proposed method enables us to restore lost pixel values.",3542
holes information typically methods Existing inpainting image from fill surrounding borrowing by pixels.,Existing image inpainting methods typically fill holes by borrowing information from surrounding pixels.,3543
processes. implement interest of services great to Web Data-intensive data-intensive are,Data-intensive Web services are of great interest to implement data-intensive processes.,3544
"two will particular, new introduce In heuristics, we i.e.","In particular, we will introduce two new heuristics, i.e.",3545
"in Longest Subsequence(LCS) crossover of operators. Common designing services, distance","Longest Common Subsequence(LCS) distance of services, in designing crossover operators.",3546
"a search services of technique local will incorporating proposed. be new Additionally, distance","Additionally, a new local search technique incorporating distance of services will be proposed.",3547
"high-dimensional to and technique be this scaled Unfortunately, such cannot as hyperspectral color images. data","Unfortunately, this technique cannot be scaled to high-dimensional data such as color and hyperspectral images.",3548
The for used filtering. resulting be means algorithm also nonlocal can,The resulting algorithm can also be used for nonlocal means filtering.,3549
NAS of the This wide tasks. limitation impedes large-scale on use,This limitation impedes the wide use of NAS on large-scale tasks.,3550
is chosen as one The best the basic architecture.,The best one is chosen as the basic architecture.,3551
process is architecture. The with help of basic the large-scale accelerated the search,The large-scale search process is accelerated with the help of the basic architecture.,3552
architecture-level we but transfer. mechanism NAS a a not propose method only What for is,What we propose is not only a NAS method but a mechanism for architecture-level transfer.,3553
a computer systems. component vision subtraction Background is of significant,Background subtraction is a significant component of computer vision systems.,3554
"object widely anomaly used in detection, tracking, It video is surveillance, etc.","It is widely used in video surveillance, object tracking, anomaly detection, etc.",3555
designed incurred A preprocessing influences reduce the by is to strategy depth sensors. from noise,A preprocessing strategy is designed to reduce the influences incurred by noise from depth sensors.,3556
the on regularization. One successful of learning is approaches consistency based in semi-supervised the,One of the successful approaches in semi-supervised learning is based on the consistency regularization.,3557
"existing Unfortunately, do targets. not quality the teacher the assess methods of","Unfortunately, existing methods do not assess the quality of the teacher targets.",3558
decoupled to further novel encourage difference. a model We framework introduce,We further introduce a novel decoupled framework to encourage model difference.,3559
improved via The is several strategy of validated proposed performance experiments. the,The improved performance of the proposed strategy is validated via several experiments.,3560
of a network. multi-scale encoder convolutional is and composed attention SAFE a scale,SAFE is composed of a multi-scale convolutional encoder and a scale attention network.,3561
"the problem it extracting explicitly from features characters. First, scale-invariant the scale tackles by","First, it explicitly tackles the scale problem by extracting scale-invariant features from the characters.",3562
"transfer of Second, scales. different the learning character it can encoding feature across","Second, it can transfer the learning of feature encoding across different character scales.",3563
divisions. modifications such very DNA over methylation general in stable many as cell Epigenetic are,Epigenetic modifications such as DNA methylation are in general very stable over many cell divisions.,3564
model Our at (i.e. deterministic molecular happening several hybrid decribes events either points,Our hybrid model decribes several molecular events either happening at deterministic points (i.e.,3565
mouse time-resolved test cells We data. our using embryonic on model stem,We test our model on mouse embryonic stem cells using time-resolved data.,3566
on with prediction. Experiments depth methods both other semantic compared are and labeling,Experiments are compared with other methods on both semantic labeling and depth prediction.,3567
the addresses safety paper assessment. of task This road,This paper addresses the task of road safety assessment.,3568
US states. our evaluated on We large images of two dataset real-world a approach from,We evaluated our approach on a large dataset of real-world images from two US states.,3569
"we alternating optimization. geometry for formulate programming linear an First, problem","First, we formulate an alternating linear programming problem for geometry optimization.",3570
\AD{We with of extensions. framework our specifications and functional a illustrate variety,\AD{We illustrate our framework with a variety of functional specifications and extensions.,3571
using concatenation proposed architecture The connection. by can and dense multi-level capture more features,The proposed architecture can capture more multi-level features by using concatenation and dense connection.,3572
is density The on vibration based probability non-Gaussian then noises the from distribution. extracted Gaussian,The non-Gaussian vibration is then extracted from Gaussian noises based on the probability density distribution.,3573
images favorable reports taken on under in illumination Most segmentation daytime semantic conditions. progress,Most progress in semantic segmentation reports on daytime images taken under favorable illumination conditions.,3574
Battery-powered IoT many now critical applications. wireless widely devices seen in are,Battery-powered wireless IoT devices are now widely seen in many critical applications.,3575
is demonstrate design. advantage used of to baseline the A our scheme,A baseline scheme is used to demonstrate the advantage of our design.,3576
to due still very challenging similar (e.g. appearances It is,It is still very challenging due to similar appearances (e.g.,3577
"via the occurs manipulation On training tasks, data. engagement the of supervised the user","On supervised tasks, the user engagement occurs via the manipulation of the training data.",3578
"algorithms unsupervised employed of on of depends heavily feature. the result the type Therefore,","Therefore, the result of unsupervised algorithms heavily depends on the type of employed feature.",3579
a system barrier. can invaluable The mitigate be such to proposed,The proposed system can be invaluable to mitigate such a barrier.,3580
diabetes. retinopathy is Diabetic complication the of most important,Diabetic retinopathy is the most important complication of diabetes.,3581
diagnosis or to retinal of blindness. loss avoid visual lesions Early helps,Early diagnosis of retinal lesions helps to avoid visual loss or blindness.,3582
"could detailed information. loses the Although input the simplify downsampling it images problem,","Although downsampling the input images could simplify the problem, it loses detailed information.",3583
"that are two The jointly ensuring are streams mutually. they optimized, enhanced","The two streams are jointly optimized, ensuring that they are enhanced mutually.",3584
"are with and MobileNet. ResNet conducted namely, CNN two The experiments architectures","The experiments are conducted with two CNN architectures namely, ResNet and MobileNet.",3585
validate proposed using scheme and data. data We in-vivo the simulated,We validate the proposed scheme using simulated data and in-vivo data.,3586
deep when to Person problem. this learning great re-ID are improvement achieves methods tackle introduced,Person re-ID achieves great improvement when deep learning methods are introduced to tackle this problem.,3587
 geometry a Orienting problem normals correctly is surface fundamental  consistently processing. and in,  Orienting surface normals correctly and consistently is a fundamental problem in geometry processing.,3588
multiple approach evaluate correctness the real-world our cloud speed We on point and of datasets.,We evaluate the correctness and speed of our approach on multiple real-world point cloud datasets.,3589
applied to methods of We time normal a clear compression speech. compare series and,We compare a series of time compression methods applied to normal and clear speech.,3590
high-dimensional small size. data sample The method proposed a for with appealing is,The proposed method is appealing for high-dimensional data with a small sample size.,3591
and The approaches advantages disadvantages. each TPV FPV have and,The FPV and TPV approaches each have advantages and disadvantages.,3592
"unified aim them learning framework. we into Therefore, combine a to deep","Therefore, we aim to combine them into a unified deep learning framework.",3593
"benchmark dataset we also this performance Besides various the dataset, with the on approaches.","Besides the dataset, we also benchmark the performance on this dataset with various approaches.",3594
"label dependent. of application is the definition However, a often","However, the definition of a label is often application dependent.",3595
granularity. We as label this to refer,We refer to this as label granularity.,3596
"efficiency labels training. data enhance Moreover, in CNN fine-grain","Moreover, fine-grain labels enhance data efficiency in CNN training.",3597
present consistent detection. object optimization stage We for single,We present consistent optimization for single stage object detection.,3598
"its choices design various we based effectiveness, state-of-the-art conduct evaluate To RetinaNet the detector. on","To evaluate its effectiveness, we conduct various design choices based on the state-of-the-art RetinaNet detector.",3599
been has control of strategy outbreaks. the unforeseen main For epidemic centuries isolation,For centuries isolation has been the main control strategy of unforeseen epidemic outbreaks.,3600
"full When implemented effective. is isolation delay, in and very without","When implemented in full and without delay, isolation is very effective.",3601
"However, flawless in seldom implementation feasible practice. is","However, flawless implementation is seldom feasible in practice.",3602
IoT for connectivity the range door wireless applications. new opens Long,Long range wireless connectivity opens the door for new IoT applications.,3603
with embedded long-distance paper interaction technologies introduces energy-constraint This devices. for,This paper introduces technologies for long-distance interaction with energy-constraint embedded devices.,3604
and platform for strategies the It management the proposes transmission. for sleepy power,It proposes sleepy strategies for power management for the platform and the transmission.,3605
Spatial resonance imaging critical magnetic (MRI). in resolution a imaging parameter is,Spatial resolution is a critical imaging parameter in magnetic resonance imaging (MRI).,3606
of serious the medical examples. more becomes degradation for training problem to due images This,This problem becomes more serious for medical images due to the degradation of training examples.,3607
"becomes deeper deal channels different discriminatively. on with and the network the subfeatures Thus, can","Thus, the network becomes deeper and can deal with the subfeatures on different channels discriminatively.",3608
distortions global first to and The employs generate local shape strategy variations.,The first strategy employs local and global distortions to generate shape variations.,3609
into get HMEs HME variations. structural online strategy sub-online to second more an decomposes The,The second strategy decomposes an online HME into sub-online HMEs to get more structural variations.,3610
these to The strategy structural variations. strategies hybrid combines both maximize shape and,The hybrid strategy combines both these strategies to maximize shape and structural variations.,3611
offline images to generated HMEs HME The recognition. converted for online are,The generated online HMEs are converted to images for offline HME recognition.,3612
These in competitive recent compared others are results to literature. reported,These results are competitive compared to others reported in recent literature.,3613
for its advertisements. Times incessant Square notorious display of is,Times Square is notorious for its incessant display of advertisements.,3614
other arise as in well. Advertisements mediums,Advertisements arise in other mediums as well.,3615
for unavoidable consumers. companies attractive are nearly and to They,They are attractive to companies and nearly unavoidable for consumers.,3616
"a of industry. important billion-dollar effectiveness to impacting video Therefore, is advertisements measuring","Therefore, measuring effectiveness of video advertisements is important to impacting a billion-dollar industry.",3617
"scenes. complex However, under task image challenging, is this quite especially vision","However, this vision task is quite challenging, especially under complex image scenes.",3618
information these by coarse prediction performance refined are structural effectively The results for improvements.,The coarse prediction results are effectively refined by these structural information for performance improvements.,3619
"and endomicroscopy In has become years, for used diagnostic purposes increasingly guidance. recent interventional","In recent years, endomicroscopy has become increasingly used for diagnostic purposes and interventional guidance.",3620
assessment. study quantitative A Mean (MOS) image Score quality Opinion confirms this also,A Mean Opinion Score (MOS) study also confirms this quantitative image quality assessment.,3621
in neural method We deep differentiable novel image propose networks. transformation for image sampling a,We propose a novel image sampling method for differentiable image transformation in deep neural networks.,3622
is neighbors. due This on to reliance strict direct their,This is due to their strict reliance on direct neighbors.,3623
formula differentiable image. transformed use the approximation as this We then for a,We then use this approximation as a differentiable formula for the transformed image.,3624
show multi-scale it. that compare approach we our with We and sampling outperform,We compare our approach with multi-scale sampling and show that we outperform it.,3625
"learning Recently, have the dominated methods field of on based deep recognition. text","Recently, methods based on deep learning have dominated the field of text recognition.",3626
"achieve large data, With most the a of number training performances. them of can state-of-the-art","With a large number of training data, most of them can achieve the state-of-the-art performances.",3627
The publicly available and dataset code will soon. be,The code and dataset will be publicly available soon.,3628
together with term manner. metric in the confusion We an adversarial train this conventional objective,We train this confusion term together with the conventional metric objective in an adversarial manner.,3629
"more a to field. diverse designs Moreover, it receptive receptive field provide block enhancement","Moreover, it designs a receptive field enhancement block to provide more diverse receptive field.",3630
"sequences However, computationally with these identifying is intensive. algorithms traditional","However, identifying these sequences with traditional algorithms is computationally intensive.",3631
faster. similarity achieves and high It RepeatMasker times is to,It achieves high similarity to RepeatMasker and is times faster.,3632
"excellent performance the scheme. simulation Additionally, results we the to present proposed extensive of validate","Additionally, we present extensive simulation results to validate the excellent performance of the proposed scheme.",3633
prior atlas using an method. based were Brain measured automated volumes,Brain volumes were measured using an atlas prior based automated method.,3634
Bayley-III. assessed the One-year with outcome was,One-year outcome was assessed with the Bayley-III.,3635
for a biomarker CHD patients as cerebral outcome. serve in Reduced im-paired volumes could,Reduced cerebral volumes in CHD patients could serve as a biomarker for im-paired outcome.,3636
"attention of visualizations. EVE cross-modal explainability the demonstrate we Finally, through","Finally, we demonstrate the explainability of EVE through cross-modal attention visualizations.",3637
relies accuracy the approaches on of facial heavily landmarks. Traditional,Traditional approaches heavily relies on the accuracy of facial landmarks.,3638
to adopted Integrating get is the prediction. final regression,Integrating regression is adopted to get the final prediction.,3639
The evaluated challenging on three benchmarks. approach proposed is,The proposed approach is evaluated on three challenging benchmarks.,3640
been code has The Github. on released,The code has been released on Github.,3641
visual neural be and particularly Deep audio recognition tasks. networks to have proven effective in,Deep neural networks have proven to be particularly effective in visual and audio recognition tasks.,3642
research based of trends. future include also analysis thorough for a on current We proposals,We also include proposals for future research based on a thorough analysis of current trends.,3643
Face home and office security. in have problems wide face applications detection and recognition,Face detection and face recognition problems have wide applications in home and office security.,3644
"can detected faces Using uncontrolled be system, in environments. this","Using this system, faces can be detected in uncontrolled environments.",3645
are used Trajectories to often applications. encode computer vision several information long-temporal in,Trajectories are often used to encode long-temporal information in several computer vision applications.,3646
"For problem, in this results superpixels. our disconnected","For our problem, this results in disconnected superpixels.",3647
position edges addition We similarity. and to for based exploit trajectory constraints in color,We exploit constraints for edges in addition to trajectory based color and position similarity.,3648
"to components are additional layers, another pooling important learning. convolutional layers of deep In","In additional to convolutional layers, pooling layers are another important components of deep learning.",3649
"However, pooling currently. methods have been for no graphs developed effective","However, no effective pooling methods have been developed for graphs currently.",3650
features and hConv receptive increasing capable fields of automatically. The is layer quickly computing,The hConv layer is capable of increasing receptive fields quickly and computing features automatically.,3651
was radar introduced the properties model describe to the previously clutter this of Although .,Although this model was previously introduced to describe the clutter properties of the radar .,3652
"the hinders performance. two in methods, which overall are current there However, the problems","However, there are two problems in the current methods, which hinders the overall performance.",3653
require such learning train and as to Common the re-sampling cost-sensitive prior-knowledge system. techniques,Common techniques such as re-sampling and cost-sensitive learning require prior-knowledge to train the system.,3654
stored information time and (WM) memory allows short be to manipulated Working scales. over,Working memory (WM) allows information to be stored and manipulated over short time scales.,3655
"in dominated system with Before assembly heredity? nucleotides, era, what pre-RNA the","Before assembly with nucleotides, in the pre-RNA era, what system dominated heredity?",3656
of competition? What was survival subject the,What was the subject of survival competition?,3657
competition? subject a Is still it of,Is it still a subject of competition?,3658
"hysteresis. is However, there the and heritability argument corroborating relationship between no","However, there is no argument corroborating the relationship between heritability and hysteresis.",3659
words. analogy on statistics and created lexical are relations context-based representations using Word,Word representations are created using analogy context-based statistics and lexical relations on words.,3660
for in Understanding are Natural the representations tasks. (NLU) Language inputs learning Word models,Word representations are inputs for the learning models in Natural Language Understanding (NLU) tasks.,3661
"sufficient. only to knowing language, not understand However, context the is","However, to understand language, knowing only the context is not sufficient.",3662
lines Reading a between of the is NLU. key component,Reading between the lines is a key component of NLU.,3663
not Embedding context enhances word are represented the the in which representation. deeper word relationships,Embedding deeper word relationships which are not represented in the context enhances the word representation.,3664
"word Map Self-Organizing it. order optimize fine-tune used is to In to the representation,","In order to fine-tune the word representation, Self-Organizing Map is used to optimize it.",3665
reference compared available We methods. with performance freely also,We also compared performance with freely available reference methods.,3666
"network best-performing anatomical the both predicting and lesion Overall, examined. was the labels DeepSCAN network","Overall, the DeepSCAN network predicting both lesion and anatomical labels was the best-performing network examined.",3667
"data widely Graph have and been learning. graph-structured studied for Convolutional Networks (GCNs) representation Recently,","Recently, Graph Convolutional Networks (GCNs) have been widely studied for graph-structured data representation and learning.",3668
"task. MGAL, on we provide network semi-supervised then for unified learning a Based","Based on MGAL, we then provide a unified network for semi-supervised learning task.",3669
MGAL the effectiveness experimental model. results of demonstrate Promising,Promising experimental results demonstrate the effectiveness of MGAL model.,3670
graphic generation. important for scene Layout is design and,Layout is important for graphic design and scene generation.,3671
critical layouts. for Accurate is alignment good,Accurate alignment is critical for good layouts.,3672
maximize to explored and were A expression. signal sequences of propeptide set diverse peptides,A diverse set of signal peptides and propeptide sequences were explored to maximize expression.,3673
source problem a   cooperative seeking paper studies the multi-vehicle networked system. via This,  This paper studies the cooperative source seeking problem via a networked multi-vehicle system.,3674
controllers effectiveness The quadratic is for the proved scalar fields. of proposed,The effectiveness of the proposed controllers is proved for quadratic scalar fields.,3675
"illustrative results. are Finally, validate the included theoretical examples to","Finally, illustrative examples are included to validate the theoretical results.",3676
computer translation is difficult Image-to-image problem vision. and in long-established a a,Image-to-image translation is a long-established and a difficult problem in computer vision.,3677
this we image-to-image In paper for model propose based translation. an adversarial,In this paper we propose an adversarial based model for image-to-image translation.,3678
generative Our conditional based probability on model adversarial approach. a works network,Our generative adversarial network based model works on a conditional probability approach.,3679
has Image-to-image proven produce than to Unsupervised translation. This better Multimodal results method,This method has proven to produce better results than Multimodal Unsupervised Image-to-image translation.,3680
of spontaneous underlying states Facial emotional Micro-expression facialexpressions. (MER) distinguishes subtle the Recognition,Facial Micro-expression Recognition (MER) distinguishes the underlying emotional states of spontaneous subtle facialexpressions.,3681
"Nevertheless, to works separate modules two due dividethem existing non-linearity. their into","Nevertheless, existing works dividethem into two separate modules due to their non-linearity.",3682
"approaches. these propose we are unknown, two parameters When","When these parameters are unknown, we propose two approaches.",3683
"under the is auspices of source the Arbor HBP. developed open software,","Arbor is open source software, developed under the auspices of the HBP.",3684
"approaches we In paper, bottom-up this competitively. still that perform show","In this paper, we show that bottom-up approaches still perform competitively.",3685
the aligned. if into keypoints are We geometrically bounding box group a they five,We group the five keypoints into a bounding box if they are geometrically aligned.,3686
that has boosted powerful performance on architecture various Cascade classic tasks. yet a is,Cascade is a classic yet powerful architecture that has boosted performance on various tasks.,3687
"segmentation cascade remains how to However, open question. to instance introduce an","However, how to introduce cascade to instance segmentation remains an open question.",3688
combination Cascade only gain. simple A brings of and R-CNN Mask R-CNN limited,A simple combination of Cascade R-CNN and Mask R-CNN only brings limited gain.,3689
"definitely errors. will this to lead recognition applying In closed-set methods case, unseen-category","In this case, applying closed-set recognition methods will definitely lead to unseen-category errors.",3690
"methods. even ODN can performance achieve Moreover, closed-set comparable to some","Moreover, ODN can even achieve comparable performance to some closed-set methods.",3691
"difficult parts avoid However, to it accurately. between is self-intersection body","However, it is difficult to avoid self-intersection between body parts accurately.",3692
"is term the First, the region. the self-intersection self-intersection defined of as volume penalty","First, the self-intersection penalty term is defined as the volume of the self-intersection region.",3693
(kernels) predetermined have shapes weights. and sparsities only and Typical learn filters,Typical filters (kernels) have predetermined shapes and sparsities and learn only weights.,3694
and fix only A feature learn sparsities. shapes extraction weights methods and few,A few feature extraction methods fix weights and learn only shapes and sparsities.,3695
and constraints learning predetermined optimal extracting These restrict features.,These predetermined constraints restrict learning and extracting optimal features.,3696
"campaign a introduced. unmanned-aerial-vehicle this channels paper, conducted is measurement for recently (UAV) In","In this paper, a recently conducted measurement campaign for unmanned-aerial-vehicle (UAV) channels is introduced.",3697
Five routes and considered. vertical five horizontal were flight,Five horizontal and five vertical flight routes were considered.,3698
in (OCT) become tomography the has important most Optical modality ophthalmology. coherence imaging,Optical coherence tomography (OCT) has become the most important imaging modality in ophthalmology.,3699
"model with images segmentation source First, we a device. on OCT a train acquired","First, we train a segmentation model on images acquired with a source OCT device.",3700
imbalanced in object appear ratios. classes Real-world,Real-world object classes appear in imbalanced ratios.,3701
which a classifiers poses challenge towards get frequent biased significant for This classes.,This poses a significant challenge for classifiers which get biased towards frequent classes.,3702
"two of underlying The i.e. requirements face progression, age","The two underlying requirements of face age progression, i.e.",3703
"literature. permanence, accuracy are in not the aging and identity studied well","aging accuracy and identity permanence, are not well studied in the literature.",3704
show assessments quantitative advances visual Both the the and that approach state-of-the-art.,Both visual and quantitative assessments show that the approach advances the state-of-the-art.,3705
"the in Second, probability evidence with annotations removed. the map are building no","Second, the annotations with no evidence in the building probability map are removed.",3706
"adversarial a using propose compressed framework generative JPEG Hence, images we generates novel networks. that","Hence, we propose a novel framework that generates JPEG compressed images using generative adversarial networks.",3707
layer to proposed is operations. enable block-based The locally connected,The locally connected layer is proposed to enable block-based operations.,3708
approach ethics. a machine to fuzzy paper reasoning This describes,This paper describes a fuzzy reasoning approach to machine ethics.,3709
method propose select clustering classes. to a the We based representative,We propose a clustering based method to select the representative classes.,3710
method proposed that results the generation Experimental the CAM obviously. show improves,Experimental results show that the proposed method improves the CAM generation obviously.,3711
"angle load for are estimating presented. the and Finally, slip tire novel algorithms","Finally, novel algorithms for estimating the tire load and slip angle are presented.",3712
Experimental effectiveness the algorithms. of proposed the results demonstrate,Experimental results demonstrate the effectiveness of the proposed algorithms.,3713
sparsity image in been utilized sensing self-similarity have widely Nonlocal and (CS). compressive group,Nonlocal self-similarity and group sparsity have been widely utilized in image compressive sensing (CS).,3714
methods two into can approaches. Existing and categorized bottom-up groups: be top-down,Existing methods can be categorized into two groups: top-down and bottom-up approaches.,3715
"among generic object candidates. Firstly, human we methods a to state-of-the-art detector detect choose","Firstly, we choose a generic object detector among state-of-the-art methods to detect human candidates.",3716
"corresponding pyramid Then, pose. cascaded the network estimate used human to the is","Then, the cascaded pyramid network is used to estimate the corresponding human pose.",3717
conduct We experiments configurations. of ablative models extensive validate and various to choices,We conduct extensive ablative experiments to validate various choices of models and configurations.,3718
modulation to is robust and approach more noise can This different techniques. with cope,This approach is more robust to noise and can cope with different modulation techniques.,3719
"statistical transient. the from of energy Then, extracted a features is set","Then, a set of statistical features is extracted from the energy transient.",3720
"to selected fed Finally, learning algorithms for several machine are classification. features","Finally, selected features are fed to several machine learning algorithms for classification.",3721
signals wirelessly are oscilloscope. a using recorded high-frequency The,The signals are recorded wirelessly using a high-frequency oscilloscope.,3722
for and tested methods results ratio reported. Proposed levels (SNR) are different are signal-to-noise also,Proposed methods are also tested for different signal-to-noise ratio (SNR) levels and results are reported.,3723
on using integral approximations and deconvolution Proposed methods direct transforms. non-blind base without,Proposed methods base on direct non-blind deconvolution without using approximations and integral transforms.,3724
be data. adapted may multi-dimensional Methods of for reconstruction,Methods may be adapted for reconstruction of multi-dimensional data.,3725
presented. comparative combined and is analysis step-by-step The shifts of methods,The comparative analysis of step-by-step and combined shifts methods is presented.,3726
and considered. restorations signal of reconstructions distortions are at different image Examples,Examples of signal reconstructions and image restorations at different distortions are considered.,3727
on of depend applications of image used. efficiency vision the algorithms computer variety A matching,A variety of computer vision applications depend on the efficiency of image matching algorithms used.,3728
features in match detect descriptors designed Various are to and images.,Various descriptors are designed to detect and match features in images.,3729
time. need Deployment in for mobile computation low this algorithms applications creates of a,Deployment of this algorithms in mobile applications creates a need for low computation time.,3730
"often order matched decrease of time In compromised. is to quality keypoints complexity,","In order to decrease time complexity, quality of keypoints matched is often compromised.",3731
infers regions or image. object for in one more an detection object Region-based categories,Region-based object detection infers object regions for one or more categories in an image.,3732
"R-DAD, into proposed In multiple the first an object decompose region we regions. small","In the proposed R-DAD, we first decompose an object region into multiple small regions.",3733
"can its You nothing imagine experience, appearance your given more.","You can imagine its appearance given your experience, nothing more.",3734
previous are Results and proposal. any other outperform impressive,Results are impressive and outperform any other previous proposal.,3735
Learning significant datasets models unbiased challenge. imbalanced is on a,Learning unbiased models on imbalanced datasets is a significant challenge.,3736
estimation as uncertainty epistemic This framework dropout retrieval. Bayesian using in a approximation enables,This enables epistemic uncertainty estimation using dropout as a Bayesian approximation framework in retrieval.,3737
"leveraged boost Monte is to Carlo (MC) Accordingly, sampling performance. retrieval","Accordingly, Monte Carlo (MC) sampling is leveraged to boost retrieval performance.",3738
car evaluated driving. approach person Our applications: two is autonomous re-identification on and,Our approach is evaluated on two applications: person re-identification and autonomous car driving.,3739
state-of-the-art on datasets Comparable are achieved application. the results multiple for former,Comparable state-of-the-art results are achieved on multiple datasets for the former application.,3740
application. autonomous driving (HDD) We the for Honda dataset leverage driving car,We leverage the Honda driving dataset (HDD) for autonomous car driving application.,3741
notions and ego-motion It provides for similarity modalities understanding. action multiple,It provides multiple modalities and similarity notions for ego-motion action understanding.,3742
"present we Hence, conditional network. a retrieval multi-modal","Hence, we present a multi-modal conditional retrieval network.",3743
representations separate disentangles It embeddings different encode to into similarities.,It disentangles embeddings into separate representations to encode different similarities.,3744
been a energy principle function. has The as free proposed brain unifying account of,The free energy principle has been proposed as a unifying account of brain function.,3745
"clarifies distinctive article predictions. and shared apart This these connections, teasing","This article clarifies these connections, teasing apart distinctive and shared predictions.",3746
image posed is transformation aligns as The problem that a given pair. geometric finding the,The problem is posed as finding the geometric transformation that aligns a given image pair.,3747
and dataset train benchmark Proposal-Flow approach test standard our (PF-PASCAL)\cite{proposal_flow}. on We a,We train and test our approach on a standard benchmark dataset Proposal-Flow (PF-PASCAL)\cite{proposal_flow}.,3748
approach proposed the The results the demonstrating of state-of-the-art method. effectiveness achieves,The proposed approach achieves state-of-the-art results demonstrating the effectiveness of the method.,3749
performance. translation the FID to is The used evaluate quantitatively,The FID is used to quantitatively evaluate the translation performance.,3750
and unsupervised of unpaired with optical also The explored. learning SAR images possibility is,The possibility of unsupervised learning with unpaired SAR and optical images is also explored.,3751
"of our the is the To best there literature. application similar documented knowledge, in no","To the best of our knowledge, there is no similar application documented in the literature.",3752
"and to scenes. specifically, More street use LiDAR cameras scan we","More specifically, we use LiDAR and cameras to scan street scenes.",3753
composite and sensor images models. viewpoints can with The be re-synthesized different,The composite images can be re-synthesized with different viewpoints and sensor models.,3754
"traditional approaches, unmatched Compared offers method scalability to and our realism.","Compared to traditional approaches, our method offers unmatched scalability and realism.",3755
these we images quantitative multiple projections obtain From angiography maps. and simultaneously,From these images we simultaneously obtain angiography projections and multiple quantitative maps.,3756
image-to-image in state-of-the-art impressive results unsupervised yielded works Several translation. have the GANs-based,Several state-of-the-art works have yielded impressive results in the GANs-based unsupervised image-to-image translation.,3757
"accurately-annotated these indispensably on achievements However, the rely datasets.","However, these achievements indispensably rely on the accurately-annotated datasets.",3758
"usefulness design samples. correctly-labeled re-weighting hard Next, the strategy preserve of a to we sample","Next, we design a sample re-weighting strategy to preserve the usefulness of correctly-labeled hard samples.",3759
"very and Our promising proposed skin results. validated on lesion method is achieved classification task,","Our proposed method is validated on skin lesion classification task, and achieved very promising results.",3760
"for two However, are treatment there method. drawbacks this","However, there are two drawbacks for this treatment method.",3761
research Biometrics-related deep  been technology. learning has significantly by accelerated ,  Biometrics-related research has been accelerated significantly by deep learning technology.,3762
commodity real-time creating video hardware. a present method summaries in for We on,We present a method for creating video summaries in real-time on commodity hardware.,3763
"discard use low-level features First, undesirable are frames. to","First, low-level features are use to discard undesirable frames.",3764
"are segment. and segment-level extracted divided segments, each Next, features for video into is","Next, video is divided into segments, and segment-level features are extracted for each segment.",3765
learning research. visual largely problem computer remains Unsupervised in representation unsolved vision a,Unsupervised visual representation learning remains a largely unsolved problem in computer vision research.,3766
classification to loss embedding a boost regularizer We as feature performance. employ triplet,We employ triplet loss as a feature embedding regularizer to boost classification performance.,3767
generality pretrained fine-tuning This promotes while networks.,This promotes generality while fine-tuning pretrained networks.,3768
recently embedding proposed is regularizers. loss powerful surrogate for Triplet a,Triplet loss is a powerful surrogate for recently proposed embedding regularizers.,3769
"requirement is high avoided to batch-size computational and due large cost. Yet, it","Yet, it is avoided due to large batch-size requirement and high computational cost.",3770
"these assumptions. we re-assess Through experiments, our","Through our experiments, we re-assess these assumptions.",3771
"network supports During tasks without embedding and both our inference, classification any overhead. computational","During inference, our network supports both classification and embedding tasks without any computational overhead.",3772
steady Quantitative recognition a five datasets. improvement on evaluation highlights fine-grained,Quantitative evaluation highlights a steady improvement on five fine-grained recognition datasets.,3773
evaluation dataset imbalanced video achieves on improvement. significant an Further,Further evaluation on an imbalanced video dataset achieves significant improvement.,3774
loss embedding brings Triplet characteristics models. to classification nearest neighbor feature like,Triplet loss brings feature embedding characteristics like nearest neighbor to classification models.,3775
of the demonstrate our Our experiments utility approach.,Our experiments demonstrate the utility of our approach.,3776
depth as Monocular is ill-posed ambiguous and often an estimation described problem. inherently,Monocular depth estimation is often described as an ill-posed and inherently ambiguous problem.,3777
contribute depth All to accurate factors in estimation. these difficulty,All these factors contribute to difficulty in accurate depth estimation.,3778
then We papers understand these made over the and improvements one another. compare,We then compare these papers and understand the improvements made over one another.,3779
"explore potential better we Finally, this aid problem. improvements can to that solve","Finally, we explore potential improvements that can aid to better solve this problem.",3780
performance? is in pre-training and the ii) effective furthering the,and ii) is the pre-training effective in furthering the performance?,3781
"question, answer the second several we carried To experiments. out ablation","To answer the second question, we carried out several ablation experiments.",3782
"systems scheme open an question. the the Hence, for best remains UL maMIMO still","Hence, the best scheme for the UL maMIMO systems still remains an open question.",3783
synthetic face effect Such recognition on have algorithms. adverse alterations,Such synthetic alterations have adverse effect on face recognition algorithms.,3784
"searching, multi-camera Each represents of end-of-track. different a and three tracking, status track: stages","Each status represents three different stages of a multi-camera track: tracking, searching, and end-of-track.",3785
tracked tracking status are a targets single camera by The tracker. means,The tracking status means targets are tracked by a single camera tracker.,3786
"other are targets examined the searching the In disappeared reappear they in if status, cameras.","In the searching status, the disappeared targets are examined if they reappear in other cameras.",3787
the target status network exited its due the invisibility. The end-of-track camera to does lengthy,The end-of-track status does the target exited the camera network due to its lengthy invisibility.,3788
to trees track-hypothesis tracking. three assists multi-camera the for status form These MHT,These three status assists MHT to form the track-hypothesis trees for multi-camera tracking.,3789
"a gating of Furthermore, they for unlikely association. observation-to-track technique present eliminating","Furthermore, they present a gating technique for eliminating of unlikely observation-to-track association.",3790
"can operate the real-time they that show proposed and in In method online. addition,","In addition, they show that the proposed method can operate in real-time and online.",3791
the of simulation verify used unified proposed the approach. are results The to effectiveness,The simulation results are used to verify the effectiveness of the proposed unified approach.,3792
takes using learning-based towards Scrabble steps self-play. a thesis building This agent,This thesis takes steps towards building a learning-based Scrabble agent using self-play.,3793
and light-weight this crucial scalable is while architectures problem. Designing addressing,Designing scalable and light-weight architectures is crucial while addressing this problem.,3794
the predict entire single reconstruction approaches point a based directly cloud point-cloud in stage. Existing,Existing point-cloud based reconstruction approaches directly predict the entire point cloud in a single stage.,3795
"that are uniform point dense. accurate, generates clouds method and Our","Our method generates point clouds that are accurate, uniform and dense.",3796
base will We our code make available. fully,We will make our code base fully available.,3797
introduce the We of scene-aware task dialog.,We introduce the task of scene-aware dialog.,3798
"Scene-Aware Dialog this the benchmark (AVSD) Dataset. To Visual introduce Audio task, we","To benchmark this task, we introduce the Audio Visual Scene-Aware Dialog (AVSD) Dataset.",3799
arbitrary Periodicity length signal noisy costly. computationally very is an estimation from,Periodicity estimation from an arbitrary length noisy signal is computationally very costly.,3800
"mask. FCNN the trained soft In is the the methodology proposed, using","In the methodology proposed, the FCNN is trained using the soft mask.",3801
"than larger of available, forgery comparable, datasets. publicly This over is an order magnitude dataset","This dataset is over an order of magnitude larger than comparable, publicly available, forgery datasets.",3802
"data, on a performed we thorough detectors. Based of data-driven forgery this analysis","Based on this data, we performed a thorough analysis of data-driven forgery detectors.",3803
Experiments PRN and perform on real rainy PReNet both that images. and favorably synthetic show,Experiments show that PRN and PReNet perform favorably on both synthetic and real rainy images.,3804
"when This e.g. issue automatic tools, is processing particular a","This is a particular issue when automatic processing tools, e.g.",3805
"later employed quantitative image segmentation methods, biomarkers are for measures derive to or analyses.","image segmentation methods, are employed to derive quantitative measures or biomarkers for later analyses.",3806
QC each visual feasible segmentation Manual and scale. of inspection isn't large at,Manual inspection and visual QC of each segmentation isn't feasible at large scale.,3807
a then into separated segment image use cooperative to classes. technique We two,We then use a cooperative technique to segment image into two separated classes.,3808
"Additionally, sum a applicable operator object system. the general to from a is extended","Additionally, the applicable object is extended from a sum operator to a general system.",3809
the introduced are methodology. illustrate the to flexibility numerical and of Three performed examples applicability,Three numerical examples are performed to illustrate the applicability and flexibility of the introduced methodology.,3810
communication. require communications and visible simultaneous illumination Indoor (VLC) light,Indoor visible light communications (VLC) require simultaneous illumination and communication.,3811
"have codes systems. Recently, output proposed multiple VLC been multiple input for (MIMO)","Recently, codes have been proposed for multiple input multiple output (MIMO) VLC systems.",3812
"for illumination dimming not considered been uniform control along However, has VLC systems. MIMO with","However, uniform illumination along with dimming control has not been considered for MIMO VLC systems.",3813
"proposed codes. We Hamming present run-length, the and the rate, distance for for code expressions","We present the expressions for code rate, run-length, and Hamming distance for the proposed codes.",3814
"in of type be problems, some the different. multiple might data views Furthermore,","Furthermore, in some problems, the data type of multiple views might be different.",3815
the proposed demonstrate studies algorithm. on Empirical real-world data sets multiple of our effectiveness Deep-MTMV,Empirical studies on multiple real-world data sets demonstrate the effectiveness of our proposed Deep-MTMV algorithm.,3816
frequency are rotation depolarization and suitable combinations Waveplate for discussed. fast sequence,Waveplate sequence and rotation frequency combinations suitable for fast depolarization are discussed.,3817
constellations signal the are optimized. all of entry size Both and,Both the size and entry of all signal constellations are optimized.,3818
"Specifically, use size we for optimization. design the suggest recursive of a","Specifically, we suggest the use of a recursive design for size optimization.",3819
"convolutional (FCNs). been semantic in Recently, networks results have fully using state-of-the-art achieved segmentation","Recently, state-of-the-art results have been achieved in semantic segmentation using fully convolutional networks (FCNs).",3820
only information truth the frameworks by maps. exploit class Such segmentation using ground,Such frameworks only exploit class information by using the ground truth segmentation maps.,3821
"different model multi-modal domains, handles from information The i.e. sources","The model handles multi-modal information sources from different domains, i.e.",3822
used bias background constraint to as is patch texturing. environmental image An,An image background patch is used as constraint to bias environmental texturing.,3823
point-cloud directly applied the is within (Point-Net). global function A generator the on approximation,A global approximation function within the generator is directly applied on the point-cloud (Point-Net).,3824
of image. viewpoint and bias the used the are generated to background Conditions the,Conditions are used to bias the background and the viewpoint of the generated image.,3825
predominantly are of action. subject primary videos humans performing The these some,The primary subject of these videos are predominantly humans performing some action.,3826
video anywhere proposals. automated range inter-action between applications from human-computer editing to The can,The applications can range from anywhere between human-computer inter-action to automated video editing proposals.,3827
"we localization. spatiotemporal action we with recognition, action deal When consider","When we consider spatiotemporal action recognition, we deal with action localization.",3828
class approach novel a for We present (CNN) based classification. one Convolutional Network Neural,We present a novel Convolutional Neural Network (CNN) based approach for one class classification.,3829
achieves proposed the recent the over Extensive experiments demonstrate methods. improvements method state-of-the-art that significant,Extensive experiments demonstrate that the proposed method achieves significant improvements over the recent state-of-the-art methods.,3830
github.com/otkupjnoz/oc-cnn. source available The code at is :,The source code is available at : github.com/otkupjnoz/oc-cnn.,3831
cells and observed heterogeneity often Phenotypic cancer in is resistance. drug to linked is widely,Phenotypic heterogeneity in cancer cells is widely observed and is often linked to drug resistance.,3832
training prototypes all The to are learnt of be data. representative,The prototypes are learnt to be representative of all training data.,3833
as an to test-time encoding classifier. we input At this a utilize,At test-time we utilize this encoding as an input to a classifier.,3834
We on conduct benchmark experiments datasets. extensive,We conduct extensive experiments on benchmark datasets.,3835
"Our outperforms specific the method DA). FSL, (ZSL, methods state-of-art contexts trained for","Our method outperforms state-of-art methods trained for the specific contexts (ZSL, FSL, DA).",3836
propagation essential. the knowledge is Accurate mm-wave and of fundamental channels,Accurate knowledge of the mm-wave propagation channels is fundamental and essential.,3837
end-to-end Convolutional for called fine-grained (CoCoNet). an We network Collaborative present deep visual categorization Network,We present an end-to-end deep network for fine-grained visual categorization called Collaborative Convolutional Network (CoCoNet).,3838
shows method proposed parts its outperforms consistently. ablation constituent The that the study,The ablation study shows that the proposed method outperforms its constituent parts consistently.,3839
outperforms CoCoNet state-of-the-art also methods. competing few,CoCoNet also outperforms few state-of-the-art competing methods.,3840
into These be network features neural taken end-to-end visual inputs. an will as next,These visual features will be next taken into an end-to-end neural network as inputs.,3841
method generate AMOS described. semiautomatic is to used Patches The,The semiautomatic method used to generate AMOS Patches is described.,3842
"clustering and selection. It selection, viewpoint patch includes camera","It includes camera selection, viewpoint clustering and patch selection.",3843
"patches. the source both we as as registered For training, the provide full well images","For training, we provide both the registered full source images as well as the patches.",3844
on changes matching benchmarks. state-of-the-art illumination achieves standard It under in,It achieves state-of-the-art in matching under illumination changes on standard benchmarks.,3845
"Networks Recently, proposed, limitation. can this help which are (CapsNets) eliminating Capsule","Recently, Capsule Networks (CapsNets) are proposed, which can help eliminating this limitation.",3846
"introduce (PS) reduce capsules. of To the parameters, Parameter-Sharing the between number we mechanism","To reduce the number of parameters, we introduce the Parameter-Sharing (PS) mechanism between capsules.",3847
the two accuracy classification The proposed in folds. improves WS-DAN,The proposed WS-DAN improves the classification accuracy in two folds.,3848
"spatiotemporal the By for our we further need labeling. human anomaly doing, without enlarge dataset","By doing, we enlarge our spatiotemporal anomaly dataset without the need for further human labeling.",3849
"network for two-stream this we a recognition. paper, present fashion In multi-task","In this paper, we present a two-stream multi-task network for fashion recognition.",3850
"have field. monocular estimation deep-convolutional-neural-networks achieved methods depth based results in promising Recently, the","Recently, deep-convolutional-neural-networks based methods have achieved promising results in the monocular depth estimation field.",3851
"to propose we this In attention-based network tackle these aggregation paper, context difficulties. (ACAN) an","In this paper, we propose an attention-based context aggregation network (ACAN) to tackle these difficulties.",3852
"recast dense classification as depth we a labeling First, monocular problem. estimation multi-class the","First, we recast the monocular depth estimation as a dense labeling multi-class classification problem.",3853
two handle Probability theory and germane information. represent and to theories Dempster-Shafer are theory uncertain,Probability theory and Dempster-Shafer theory are two germane theories to represent and handle uncertain information.,3854
in in images satellite first-step many remote an detection sensing important is applications. Cloud,Cloud detection in satellite images is an important first-step in many remote sensing applications.,3855
are This limited problem more spectral available. only is number a of bands challenging when,This problem is more challenging when only a limited number of spectral bands are available.,3856
"To is this in algorithm deep proposed paper. address a problem, learning-based this","To address this problem, a deep learning-based algorithm is proposed in this paper.",3857
"proposed an end-to-end is solution, no required. the step method pre-processing is Since complicated","Since the proposed method is an end-to-end solution, no complicated pre-processing step is required.",3858
extensively fitting perspectives. random been has and studied the from Multi-model sampling clustering,Multi-model fitting has been extensively studied from the random sampling and clustering perspectives.,3859
"are In words, other network. embedded clusters through of together closer points same the the","In other words, points of the same clusters are embedded closer together through the network.",3860
fitting multi-model promising with well. tasks as made on also Comparisons are single-type results,Comparisons are also made on single-type multi-model fitting tasks with promising results as well.,3861
orientation voxel-wise Each one of orientation the principal tract. tract map represents,Each tract orientation map represents the voxel-wise principal orientation of one tract.,3862
with automatic extensions These bundle-specific accuracy. creation the unseen of previously enable tractograms,These extensions enable the automatic creation of bundle-specific tractograms with previously unseen accuracy.,3863
for CVNN is a channel as single with a entity. complex-valued dealing communication suitable fading,CVNN is suitable for dealing with a fading communication channel as a single complex-valued entity.,3864
"Hence, accuracy. limited stationary gives an prediction only empirically network selected","Hence, an empirically selected stationary network gives only limited prediction accuracy.",3865
We simulations its wireless-propagation in real online adaptability experiments. and demonstrate,We demonstrate its online adaptability in simulations and real wireless-propagation experiments.,3866
"network training accelerates conventional with significantly also methods. RLD deep compared Furthermore,","Furthermore, RLD also significantly accelerates deep network training compared with conventional methods.",3867
"under In present HCV this analysis therapy. the we a paper, of model global","In this paper, we present the global analysis of a HCV model under therapy.",3868
of SDTS-based our Experimental demonstrate effectiveness the method. results,Experimental results demonstrate the effectiveness of our SDTS-based method.,3869
risk. express phrases often estimated Verbal to are used probability,Verbal probability phrases are often used to express estimated risk.,3870
studied these before. of been not phrases Many had,Many of these phrases had not been studied before.,3871
The presented context were in situations. phrases the ordinary of,The phrases were presented in the context of ordinary situations.,3872
with was native their statisticians Dutch distributed survey The non-statisticians both language. as and among,The survey was distributed among both statisticians and non-statisticians with Dutch as their native language.,3873
"Furthermore, Dutch the asymmetry an interpretation of in demonstrated the results phrases. complementary","Furthermore, the results demonstrated an asymmetry in the interpretation of Dutch complementary phrases.",3874
"may Therefore, for be miscommunication. expressions probability a risk verbal","Therefore, verbal probability expressions may be a risk for miscommunication.",3875
The accurate its landmark to estimation learns network how estimate is.,The network learns to estimate how accurate its landmark estimation is.,3876
experiments the evaluate the correlation validation The last accuracy. signal of the landmark with,The last experiments evaluate the correlation of the validation signal with the landmark accuracy.,3877
experiments three performed datasets. All were for all,All experiments were performed for all three datasets.,3878
apply Ordered Weighted operator decision-making. multi-criteria the We in Averaging (OWA),We apply the Ordered Weighted Averaging (OWA) operator in multi-criteria decision-making.,3879
"now. until system in has However, idea fuzzy been not this applied","However, this idea has not been applied in fuzzy system until now.",3880
the of We order discuss review number. mechanism the and fuzzy OWA concept operator of,We review the concept of OWA operator and discuss the order mechanism of fuzzy number.,3881
of an with score measure-based give alternatives. to overall we each combine Then dominance,Then we combine with measure-based dominance to give an overall score of each alternatives.,3882
is the procedure. illustrated whole An show example to,An example is illustrated to show the whole procedure.,3883
"rock on the of of study many micro-level earth In the problem arises. branches sciences,","In many branches of earth sciences, the problem of rock study on the micro-level arises.",3884
"number samples significant feasible. is a not always representative However, of","However, a significant number of representative samples is not always feasible.",3885
generation the of similar of problem properties with actual. becomes samples the Thus,Thus the problem of the generation of samples with similar properties becomes actual.,3886
experiments good provides Minkowski of functionals. this reconstruction method terms in that show Numerical,Numerical experiments show that this method provides good reconstruction in terms of Minkowski functionals.,3887
"presents learning itself an as important tool. transfer Hence,","Hence, transfer learning presents itself as an important tool.",3888
space? analyze to the how But resulting data,But how to analyze the resulting data space?,3889
deep learning a How of traditional in is use segmentation the setting? important methods,How important is the use of traditional segmentation methods in a deep learning setting?,3890
"hand, needs stereo memory. On Block less much the matching other","On the other hand, Block matching stereo needs much less memory.",3891
our We on tested stereo the benchmark. block multi matching new Middlebury scheme,We tested our new multi block matching scheme on the Middlebury stereo benchmark.,3892
"person network, net, proposed an for development. dependent A AlterEgo is called","A person dependent network, called an AlterEgo net, is proposed for development.",3893
dialogue a with a The of person's representation network the simulates object.,The network simulates a person's dialogue with a representation of the object.,3894
a iterations 'internal stream a of the human. world' The representing generate 'forward',The 'forward' iterations generate a stream representing the 'internal world' of a human.,3895
a representation internal object. representing stream generate iterations an of the The 'backward',The 'backward' iterations generate a stream representing an internal representation of the object.,3896
a for these nets technology Thereby new augmentation is human various potentially applications.,Thereby these nets is potentially a new human augmentation technology for various applications.,3897
non-overlapping to match re-identification views. across identities camera aims (Re-ID) Person,Person re-identification (Re-ID) aims to match identities across non-overlapping camera views.,3898
proposed cross-view pairwise Re-ID models many supervised labelled data. have of quantities require which Researchers,Researchers have proposed many supervised Re-ID models which require quantities of cross-view pairwise labelled data.,3899
"representation In jointly the learns and the unsupervised feature metric. asymmetric way, DECAMEL such a","In such a way, DECAMEL jointly learns the feature representation and the unsupervised asymmetric metric.",3900
X-ray consider previous left-hand image works their in a works. All child the of,All the previous works consider left-hand X-ray image of a child in their works.,3901
"services. data, of exchanged between large be need technological by quantities produced advances, Additionally, to","Additionally, large quantities of data, produced by technological advances, need to be exchanged between services.",3902
Evolutionary fully-automated Researchers Computing factors. (EC) all the have proposed techniques above WSC meet that,Researchers have proposed Evolutionary Computing (EC) fully-automated WSC techniques that meet all the above factors.,3903
"are those or works effective. However, efficient not","However, those works are not efficient or effective.",3904
"comparing datasets methods. is using evaluation An out, state-of-the-art carried existing benchmark","An evaluation using benchmark datasets is carried out, comparing existing state-of-the-art methods.",3905
the forecasting given provides Meteorological weather time. reliable a of future within prediction about interval,Meteorological forecasting provides reliable prediction about the future weather within a given interval of time.,3906
"MeteoLOG several on rests logics. on probabilistic and traditions, temporal fuzzy, mainly","MeteoLOG rests on several traditions, mainly on fuzzy, temporal and probabilistic logics.",3907
example a finally world that forecasting propose We models real an scenario.,We finally propose an example that models a real world forecasting scenario.,3908
segmentation. topological of of numbers the The consists knowledge the Betti sequence prior of desired,The topological prior knowledge consists of the sequence of desired Betti numbers of the segmentation.,3909
long standing Face in recognition field the of is a (AI). Artificial Intelligence challenge,Face recognition is a long standing challenge in the field of Artificial Intelligence (AI).,3910
"detect, recognize, understand and goal accurately The create is human faces. verify, systems to that","The goal is to create systems that accurately detect, recognize, verify, and understand human faces.",3911
about Every reflects us. something face unique,Every face reflects something unique about us.,3912
every face recognition accurately work expect We equally for to face.,We expect face recognition to work equally accurately for every face.,3913
schemes using well-established the from scientific generated coding ten The facial are annotations literature.,The annotations are generated using ten well-established facial coding schemes from the scientific literature.,3914
measures schemes provide features. facial coding human-interpretable of facial quantitative The,The facial coding schemes provide human-interpretable quantitative measures of facial features.,3915
publicly and pretrained are The code source models available.,The source code and pretrained models are publicly available.,3916
aggressive metastasis liver liver one of malignancies. most is Colorectal,Colorectal liver metastasis is one of most aggressive liver malignancies.,3917
dramatically performance. the enhance that can Experimental consistent viewpoints results the demonstrate,Experimental results demonstrate that the consistent viewpoints can dramatically enhance the performance.,3918
the addressing and There applications are works MC methodologies. plenty of behaviors of,There are plenty of works addressing the behaviors and applications of MC methodologies.,3919
from review the signal a This approaches for MC of perspective work comprehensive provides processing.,This work provides a comprehensive review for MC approaches from the perspective of signal processing.,3920
"solving types Next, are reviewed. algorithms optimization of MC four representative the problem","Next, four representative types of optimization algorithms solving the MC problem are reviewed.",3921
"application fields evaluated. described MC Ultimately, three and are of different","Ultimately, three different application fields of MC are described and evaluated.",3922
discuss the phenomenon. behind intuition We this,We discuss the intuition behind this phenomenon.,3923
a Parameters networks neural of memory. of amount require huge recent,Parameters of recent neural networks require a huge amount of memory.,3924
when learning parameters to machine These networks tasks by inputs. neural used perform are processing,These parameters are used by neural networks to perform machine learning tasks when processing inputs.,3925
a through We invasion model thin of continuum cell consider mechanical membranes.,We consider a continuum mechanical model of cell invasion through thin membranes.,3926
applications. in mobile led embedded success has their real-time deployment for This and systems to,This success has led to their deployment in mobile and embedded systems for real-time applications.,3927
"a principled co-design these we approach. challenges, develop To address","To address these challenges, we develop a principled co-design approach.",3928
multiplexing a of Spatial different is common beam. streams in feeder performed data,Spatial multiplexing of different data streams is performed in a common feeder beam.,3929
"of user possible. design particular Under separation multiple spatial effective constraints, is the signals","Under particular design constraints, effective spatial separation of the multiple user signals is possible.",3930
gains in performance results of the Simulation throughput. terms illustrate sum,Simulation results illustrate the performance gains in terms of sum throughput.,3931
"training require collect However, abundant data to change. these environmental from suffer approaches and","However, these approaches require to collect abundant training data and suffer from environmental change.",3932
sensors. the The extraction of the on relies azimuth physical of properties PIR change,The extraction of the azimuth change relies on the physical properties of PIR sensors.,3933
"change. is environmental and abundant system robust to no are Therefore, needed training the data","Therefore, no abundant training data are needed and the system is robust to environmental change.",3934
require amount efficient This coding data to leads representations that huge schemes. image of with,This leads to image representations with huge amount of data that require efficient coding schemes.,3935
"into In sub-aperture lenslet paper, images. images this rendered are","In this paper, lenslet images are rendered into sub-aperture images.",3936
organized pseudo-sequence as images video These input HEVC the for a are codec.,These images are organized as a pseudo-sequence input for the HEVC video codec.,3937
a spatial the pixel Most trade-off algorithms color between and compute level. superpixel features at,Most superpixel algorithms compute a trade-off between spatial and color features at the pixel level.,3938
"new this method. we In Texture-Aware a issues SuperPixel these with address (TASP) paper,","In this paper, we address these issues with a new Texture-Aware SuperPixel (TASP) method.",3939
The comparative framework. an performed within image study retrieval is,The comparative study is performed within an image retrieval framework.,3940
evaluate in accuracy. attributes of We these terms retrieval,We evaluate these attributes in terms of retrieval accuracy.,3941
"we shortcoming, object exploit To address propose to to image semantics pixelated colorization. guide this","To address this shortcoming, we propose to exploit pixelated object semantics to guide image colorization.",3942
"proposed neural the Specifically, convolutional two network branches. includes","Specifically, the proposed convolutional neural network includes two branches.",3943
"what colors. learns other while object branch learns One the the is, the object branch","One branch learns what the object is, while the other branch learns the object colors.",3944
is scientific variety the and essential of for dynamics Monitoring network's maintenance purposes. a,Monitoring the network's dynamics is essential for a variety of maintenance and scientific purposes.,3945
"signals complex network. we on study nonlinear Here, dynamic fixed a graph","Here, we study nonlinear dynamic graph signals on a fixed complex network.",3946
matrix. linearized a dynamics We the networked interpret firstly into,We firstly interpret the networked dynamics into a linearized matrix.,3947
image. of a the This deals given memorability with the paper of prediction,This paper deals with the prediction of the memorability of a given image.,3948
"to four refers face anger, recognition The neutral. smile, and expressions: surprise,","The recognition refers to four face expressions: smile, surprise, anger, and neutral.",3949
"high of as raw/CNN the F-score geometric/CNN For geometric/SVM advantage over observed, is well. and","For F-score the high advantage of raw/CNN over geometric/CNN and geometric/SVM is observed, as well.",3950
"show differentiate that the subjects. can FNs, morphospace tasks metrics this defining cognitive and We","We show that the metrics defining this morphospace can differentiate FNs, cognitive tasks and subjects.",3951
"optimizations implementation the and of This design, paper CoreNEURON. presents","This paper presents the design, implementation and optimizations of CoreNEURON.",3952
(ILS) Switch We the present Invisible address both aspects. Light to,We present the Invisible Light Switch (ILS) to address both aspects.,3953
energy So them. is saving the invisible to,So the energy saving is invisible to them.,3954
"data JPEG this a images. lossless propose we hiding paper, scheme in In","In this paper, we propose a lossless data hiding scheme in JPEG images.",3955
proposed larger the works. than capacity state-of-the-art method has that Experimental results show a,Experimental results show that the proposed method has a larger capacity than state-of-the-art works.,3956
cellular physiology. hallmark is a of Phenotypic variation,Phenotypic variation is a hallmark of cellular physiology.,3957
"Metabolic underpins such as heterogeneity, single-cell variability. phenomena in tolerance growth microbial drug particular, and","Metabolic heterogeneity, in particular, underpins single-cell phenomena such as microbial drug tolerance and growth variability.",3958
populations metabolically two into The analysis more to subpopulations. or clonal predicts split distinct,The analysis predicts clonal populations to split into two or more metabolically distinct subpopulations.,3959
within speed sparsity delay-Doppler the domain. The limits in measurement results show urban high,The measurement results within urban speed limits show high sparsity in the delay-Doppler domain.,3960
deep presented. of classifiers for is technique novel image A learning,A novel technique for deep learning of image classifiers is presented.,3961
retrieval. been have used neural widely networks in content-based image Convolutional,Convolutional neural networks have been widely used in content-based image retrieval.,3962
on improves retrieval the performance. that Experiments fine-level two method show our datasets,Experiments on two datasets show that our method improves the fine-level retrieval performance.,3963
"achieves Meanwhile, hierarchical retrieval. results model terms in our of state-of-the-art","Meanwhile, our model achieves state-of-the-art results in terms of hierarchical retrieval.",3964
trained network. a and then are successively These bands separately through hierarchical combined split synthesis,These split bands are then trained separately and combined successively through a hierarchical synthesis network.,3965
against observations. Process-oriented cognition time-ordered must evaluated be of theories,Process-oriented theories of cognition must be evaluated against time-ordered observations.,3966
can estimated subjects. results parameters model reliably that be for Our individual indicate,Our results indicate that model parameters can be estimated reliably for individual subjects.,3967
"systems video dramatically is the surveillance. increased commercial re-ID in Hence, security deploying when danger","Hence, the security danger is dramatically increased when deploying commercial re-ID systems in video surveillance.",3968
reveal effects in the re-ID systems. clearly adversarial Comprehensive experiments,Comprehensive experiments clearly reveal the adversarial effects in re-ID systems.,3969
learning deep Lack models difficult. expert training large MR of makes annotated datasets,Lack of large expert annotated MR datasets makes training deep learning models difficult.,3970
"ratio. computed volume using was Dice (DSC), similarity coefficient metrics, Segmentation Hausdroff accuracy and distance","Segmentation accuracy was computed using Dice similarity coefficient (DSC), Hausdroff distance metrics, and volume ratio.",3971
"assumptions, include to the optimal on Based traffic. interfering mild extended readily algorithm is eco-driving","Based on mild assumptions, the optimal eco-driving algorithm is readily extended to include interfering traffic.",3972
"views between angle known. the relative to the In is both rotation assumed cases, be","In both cases, the relative rotation angle between the views is assumed to be known.",3973
"derive then is efficiently the The Gr\""{o}bner to used technique solutions. basis","The Gr\""{o}bner basis technique is then used to efficiently derive the solutions.",3974
the improves solution. Our solver existing for significantly regular cameras state-of-the-art first,Our first solver for regular cameras significantly improves the existing state-of-the-art solution.,3975
novel. generalized solver cameras is for second The,The second solver for generalized cameras is novel.,3976
"scanned diagrams the been have image stored P&ID as and files. generated, Over manually years,","Over the years, P&ID diagrams have been manually generated, scanned and stored as image files.",3977
There diagrams. challenging P&ID problems several vision with digitizing real associated are world,There are several challenging vision problems associated with digitizing real world P&ID diagrams.,3978
"in textual come world information. contain Real and several often resolutions, P&IDs noisy different","Real world P&IDs come in several different resolutions, and often contain noisy textual information.",3979
appropriate pipeline. with followed is by association detected the This the components of,This is followed by association of the detected components with the appropriate pipeline.,3980
is attention videos. purpose The better for understand visual to,The purpose is to better understand visual attention for videos.,3981
intelligence coming learning and machine artificial systems. are with increasingly Humans into contact,Humans are increasingly coming into contact with artificial intelligence and machine learning systems.,3982
"and social as accountability, of responsibility interpretability, further such We issues transparency. fairness, that argue","We further argue that issues of social responsibility such as fairness, accountability, interpretability, and transparency.",3983
have disease recent in diagnosis substantially increased Celiac prevalence and years.,Celiac disease prevalence and diagnosis have increased substantially in recent years.,3984
gold current standard of for celiac mucosal confirmation biopsies. visual The duodenal examination disease is,The current gold standard for celiac disease confirmation is visual examination of duodenal mucosal biopsies.,3985
and potential great has of diagnosis. for celiac the This accuracy disease technology improving efficiency,This technology has great potential for improving the accuracy and efficiency of celiac disease diagnosis.,3986
Local Lead Blurred Bad Minima . Images to,Blurred Images Lead to Bad Local Minima .,3987
"in To estimation propose we frames\hl{. \hl{exploit} people to problem, detecting} address pose this \hl{for","To address this problem, we propose to \hl{exploit} pose estimation \hl{for detecting} people in frames\hl{.",3988
to effective user guides the robot} This capture the follow videos. and,This guides the robot} to follow the user and capture effective videos.,3989
from We other to identification senior distinguish people. person target a use,We use person identification to distinguish a target senior from other people.,3990
are that small of potential. the EADs an plateau fluctuations action during voltage occur,EADs are small voltage fluctuations that occur during the plateau of an action potential.,3991
"Although ventricular tachycardia. tissue-level lead to including can single-cell arrhythmias, behavior, a EADs","Although a single-cell behavior, EADs can lead to tissue-level arrhythmias, including ventricular tachycardia.",3992
"can generate trained to be model In noisy from clean a such versions. documents scenarios,","In such scenarios, a model can be trained to generate clean documents from noisy versions.",3993
use versions to GANs This noisy generate of the documents. the explores of paper denoised,This paper explores the use of GANs to generate denoised versions of the noisy documents.,3994
"anatomy of First, dissections. and results present the we examine valve original","First, we examine valve anatomy and present the results of original dissections.",3995
the display valve. anatomy fiber of and mitral structure the on These information gross,These display the gross anatomy and information on fiber structure of the mitral valve.,3996
fiber this incorporate We the model. topology specify dissections the to from of information,We incorporate information from the dissections to specify the fiber topology of this model.,3997
supporting load. achieves We assume valve static pressure a while equilibrium the mechanical,We assume the valve achieves mechanical equilibrium while supporting a static pressure load.,3998
the model. The to pressurized the of solution the valve resulting configuration differential determines equations,The solution to the resulting differential equations determines the pressurized configuration of the valve model.,3999
"safely objects move. that system images detection skip can of frames Using approach, this barely","Using this approach, system can safely skip detection of images frames that objects barely move.",4000