File size: 4,446 Bytes
c00d17c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 |
---
license: mit
language:
- en
tags:
- biology
- CV
- images
- animals
- lepidoptera
- butterflies
- detection
- heliconius
- forewings
- hindwings
- separated wings
- full body
- butterfly
- RGB
- ruler
- whitebalance
- label
- colorchecker
---
## Model Card for butterfly_detection_yolo
This model takes in images of butterflies as photographed for museum collections and detects butterfly components (L/R forewings, L/R hindwings and body) as well as color checkers and metadata labels.
The detection model described here is used in the repository https://github.com/Imageomics/wing-segmentation to detect components and use Meta's Segment-Anything (SAM) model for segmentation of components.
## Model Details
yolo_detection_8m_shear_10.0_scale_0.5_translate_0.1_fliplr_0.0_best.pt is the butterfly detection model.
The yolo v8 detection model was trained on a dataset of 800 total images from the [Heliconius Collection-Cambridge Butterfly](imageomics/Heliconius-Collection_Cambridge-Butterfly), OM_STRI, and Monteiro datasets. The model uses the pretrained yolov8m.pt model.
## Model Description
The model is responsible for taking an input image (RGB) and generating bounding boxes for all classes below that are found in the image. Data augmentations applied during training include shear (10.0), scale (0.5), and translate (0.1). The model was trained for 50 epochs with an image size of 256. Note that despite defining an image size of 256, the normalized masks predicted by yolo can be rescaled to the original image size.
### Segmentation Classes
[`pixel class`] corresponding category
- [0] background
- [1] right_forewing
- [2] left_forewing
- [3] right_hindwing
- [4] left_hindwing
- [5] ruler
- [6] white_balance
- [7] label
- [8] color_card
- [9] body
### Details
model.train(data=YAML,
imgsz=256,
epochs=50,
batch=16,
device=DEVICE,
optimizer='auto',
verbose=True,
val=True,
shear=10.0,
scale=0.5,
translate=0.1,
fliplr = 0.0
)
## Metrics
Class Images Instances Box(P R mAP50 mAP50-95)
all 64 358 0.979 0.887 0.919 0.877
background 64 3 1 0 0.315 0.169
right_forewing 64 58 0.995 0.983 0.986 0.977
left_forewing 64 51 0.975 1 0.985 0.982
right_hindwing 64 59 0.997 0.966 0.993 0.977
left_hindwing 64 50 0.975 1 0.993 0.98
ruler 64 31 0.951 1 0.995 0.952
white_balance 64 18 0.984 1 0.995 0.995
label 64 50 0.996 1 0.995 0.935
color_card 64 24 0.988 1 0.995 0.992
body 64 14 0.928 0.921 0.939 0.815
**Developed by:** Michelle Ramirez
## How to Get Started with the Model
To view applications of how to load in the model file and predict masks on images, please refer to [this github repository](https://github.com/Imageomics/wing-segmentation)
## Citation
**BibTeX:**
```
@software{Ramirez_Lepidoptera_Wing_Segmentation_2024,
author = {Ramirez, Michelle},
doi = {10.5281/zenodo.10869579},
month = mar,
title = {{Lepidoptera Wing Segmentation}},
url = {https://github.com/Imageomics/wing-segmentation},
version = {1.0.0},
year = {2024}
}
```
**APA:**
Ramirez, M. (2024). Lepidoptera Wing Segmentation (Version 1.0.0) [Computer software]. https://doi.org/10.5281/zenodo.10869579
## Acknowledgements
The [Imageomics Institute](https://imageomics.org) is funded by the US National Science Foundation's Harnessing the Data Revolution (HDR) program under [Award #2118240](https://www.nsf.gov/awardsearch/showAward?AWD_ID=2118240) (Imageomics: A New Frontier of Biological Information Powered by Knowledge-Guided Machine Learning). Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.
|