Update README.md
Browse files
README.md
CHANGED
@@ -1,9 +1,6 @@
|
|
1 |
-
---
|
2 |
-
license: other
|
3 |
-
---
|
4 |
# DIVOTrack: A Novel Dataset and Baseline Method for Cross-View Multi-Object Tracking in DIVerse Open Scenes
|
5 |
|
6 |
-
This repository contains the details of dataset and the Pytorch implementation of Baseline Method CrossMOT of the Paper:
|
7 |
[DIVOTrack: A Novel Dataset and Baseline Method for Cross-View Multi-Object Tracking in DIVerse Open Scenes](https://arxiv.org/abs/2302.07676)
|
8 |
|
9 |
|
@@ -14,20 +11,12 @@ Cross-view multi-object tracking aims to link objects between frames and camera
|
|
14 |
- **<a href="#des"> <u>Dataset Description</u>**</a>
|
15 |
- **<a href="#str"> <u>Dataset Structure</u>**</a>
|
16 |
- **<a href="#dow"> <u>Dataset Downloads</u>**</a>
|
17 |
-
- **<a href="#det"> <u>Training Detector</u>**</a>
|
18 |
-
- **<a href="#sin"> <u>Single-view Tracking</u>**</a>
|
19 |
-
- **<a href="#cro"> <u>Cross-view Tracking</u>**</a>
|
20 |
- **<a href="#ref"> <u>Reference</u>**</a>
|
21 |
- **<a href="#con"> <u>Contact</u>**</a>
|
22 |
|
23 |
-
The test result of the cross-view MOT baseline method *MvMHAT* on the DIVOTrack.
|
24 |
-
![test.gif](asset/test.gif)
|
25 |
-
|
26 |
-
The ground truth of the DIVOTrack.
|
27 |
-
![gt.gif](asset/gt.gif)
|
28 |
|
29 |
## <a id="des">Dataset Description</a>
|
30 |
-
We collect data in 10 different real-world scenarios, named: `'Circle', 'Shop', 'Moving', 'Park', 'Ground', 'Gate1', 'Floor', 'Side', 'Square', 'Gate2'`. All
|
31 |
the sequences are captured by using 3 moving cameras: `'View1', 'View2', 'View3'` and are manually synchronized.
|
32 |
|
33 |
In the old version, the corresponding scenarios named: `'circleRegion', 'innerShop', 'movingView', 'park', 'playground', 'shopFrontGate', 'shopSecondFloor', 'shopSideGate', 'shopSideSquare', 'southGate'`. The corresponding camera named: `'Drone', 'View1', 'View2'`.
|
@@ -54,36 +43,7 @@ DIVOTrack
|
|
54 |
|
55 |
```
|
56 |
### <a id="dow">Dataset Downloads</a>
|
57 |
-
The whole dataset can
|
58 |
-
|
59 |
-
## <a id="det">Training Detector</a>
|
60 |
-
The training process of our detector is in `./Training_detector/` and the details can see from [Training_detector/README.md](https://github.com/shengyuhao/DIVOTrack/tree/main/Training_Detector#readme).
|
61 |
-
## <a id="sin">Single-view Tracking</a>
|
62 |
-
We conducted experiments on DIVOTrack in fix benchmarks:
|
63 |
-
|
64 |
-
| Benchmark | HOTA β | IDF1 β | MOTA β | MOTP β | MT β | ML β | AssA β | IDSw β | FM β |
|
65 |
-
| --------- | ------ | ------ | ------ | ------ | ---- | ---- | ------ | ------ | ---- |
|
66 |
-
| [DeepSort](./Single_view_Tracking/Deepsort/) | 54.3 | 59.9 | 79.6 | 81.2 | 462 | 50 | 45.0 | 1,920 | **2,504** |
|
67 |
-
| [CenterTrack](./Single_view_Tracking/CenterTrack/) | 55.3 | 62.2 | 73.4 | 80.6 | **534** | 35 | 49.2 | 1,631 | 2,950 |
|
68 |
-
| [Tracktor](./Single_view_Tracking/Tracktor/) | 48.4 | 56.2 | 66.6 | 80.8 | 517 | **22** | 40.3 | 1,382 | 3,337 |
|
69 |
-
| [FairMOT](./Single_view_Tracking/FairMOT/) | 65.3 | 78.2 | 82.7 | 81.9 | 486 | 48 | **62.7** | 731 | 3,498 |
|
70 |
-
| [TraDeS](./Single_view_Tracking/TraDeS/) | 58.9 | 67.3 | 74.2 | **82.3** | 504 | 38 | 54.0 | 1,263 | 2,647 |
|
71 |
-
|
72 |
-
Each single-view tracking baseline is evaluated using [TrackEval](https://github.com/shengyuhao/DIVOTrack/tree/main/TrackEval#readme).
|
73 |
-
|
74 |
-
## <a id="cro">Cross-view Tracking</a>
|
75 |
-
We conducted experiments on the DIVOTrack dataset using six benchmarks as well as our proposed method [CrossMOT](./CrossMOT/)
|
76 |
-
| Benchmark | CVMA β | CVIDF1 β |
|
77 |
-
| --------- | ------ | -------- |
|
78 |
-
| [OSNet](./Cross_view_Tracking/OSNet/) | 34.3 | 46.0 |
|
79 |
-
| [Strong](./Cross_view_Tracking/StrongReID/) | 40.9 | 45.9 |
|
80 |
-
| [AGW](./Cross_view_Tracking/AGW/) | 57.0 | 56.8 |
|
81 |
-
| [MvMHAT](./Cross_view_Tracking/MvMHAT/) | 61.0 | 62.6 |
|
82 |
-
| [CT](./Cross_view_Tracking/CT/) | 64.9 | 65.0 |
|
83 |
-
| [MGN](./Cross_view_Tracking/MGN/) | 33.5 | 39.4 |
|
84 |
-
| [CrossMOT](./CrossMOT/) | **72.4** | **71.1** |
|
85 |
-
|
86 |
-
With the exception of CrossMOT, all of the other Re-ID methods require [Multi_view_Tracking](https://github.com/shengyuhao/DIVOTrack/tree/main/Multi_view_Tracking#readme) to predict the tracking results after they are obtained. Finally, the results of CVMA and CVIDF1 are obtained through [MOTChallengeEvalKit_cv_test](https://github.com/shengyuhao/DIVOTrack/tree/main/MOTChallengeEvalKit_cv_test#readme).
|
87 |
|
88 |
|
89 |
## <a id="ref">Reference</a>
|
|
|
|
|
|
|
|
|
1 |
# DIVOTrack: A Novel Dataset and Baseline Method for Cross-View Multi-Object Tracking in DIVerse Open Scenes
|
2 |
|
3 |
+
This repository contains the details of the dataset and the Pytorch implementation of the Baseline Method CrossMOT of the Paper:
|
4 |
[DIVOTrack: A Novel Dataset and Baseline Method for Cross-View Multi-Object Tracking in DIVerse Open Scenes](https://arxiv.org/abs/2302.07676)
|
5 |
|
6 |
|
|
|
11 |
- **<a href="#des"> <u>Dataset Description</u>**</a>
|
12 |
- **<a href="#str"> <u>Dataset Structure</u>**</a>
|
13 |
- **<a href="#dow"> <u>Dataset Downloads</u>**</a>
|
|
|
|
|
|
|
14 |
- **<a href="#ref"> <u>Reference</u>**</a>
|
15 |
- **<a href="#con"> <u>Contact</u>**</a>
|
16 |
|
|
|
|
|
|
|
|
|
|
|
17 |
|
18 |
## <a id="des">Dataset Description</a>
|
19 |
+
We collect data in 10 different real-world scenarios, named: `'Circle', 'Shop', 'Moving', 'Park', 'Ground', 'Gate1', 'Floor', 'Side', 'Square', and 'Gate2'`. All
|
20 |
the sequences are captured by using 3 moving cameras: `'View1', 'View2', 'View3'` and are manually synchronized.
|
21 |
|
22 |
In the old version, the corresponding scenarios named: `'circleRegion', 'innerShop', 'movingView', 'park', 'playground', 'shopFrontGate', 'shopSecondFloor', 'shopSideGate', 'shopSideSquare', 'southGate'`. The corresponding camera named: `'Drone', 'View1', 'View2'`.
|
|
|
43 |
|
44 |
```
|
45 |
### <a id="dow">Dataset Downloads</a>
|
46 |
+
The whole dataset can be downloaded from [GoogleDrive](https://drive.google.com/drive/folders/1RCk95TdFv3Tt7gVuyxJasiHG1IPE6jkX?usp=sharing). **Note that, each file needs to unzip by the password. You can decompress each `.zip` file in its folder after send us ([email protected], [email protected]) the License in any format.** After that, you should run `generate_ini.py` to generate `seqinfo.ini` file.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
47 |
|
48 |
|
49 |
## <a id="ref">Reference</a>
|