CAMO-MOT
This is the official repo release of the paper CAMO-MOT: Combined Appearance-Motion Optimization for 3D Multi-Object Tracking with Camera-LiDAR Fusion.
News
- 2022-09-08. CAMO-MOT is released on arXiv:slightly_smiling_face:.
- 2022-08-04. We rank first among all methods on nuScenes Dataset for Tracking:blush:.
- 2022-08-03. We rank 4th among all methods on KITTI Dataset for MOT:grinning:.
Results
Multi-object tracking(on nuScenes test set)
| Method | AMOTA | AMOTP |
|---|---|---|
| CAMO-MOT | 0.753 | 0.472 |
You can find detailed results on nuScenes test set on this website. Or you can view the accuracy trend of MOT algorithms on this website
Multi-object tracking(on nuScenes val set)
| Tracker | AMOTA | AMOTP |
|---|---|---|
| CAMO-MOT | 0.763 | 0.527 |
On nuScenes, we use BEVFusion and FocalConv as our detectors.
Multi-object tracking(on KITTI test)
| Category | HOTA (%) | MOTA (%) | MOTP (%) | MT (%) | ML (%) | IDS | FRAG | FP | FN |
|---|---|---|---|---|---|---|---|---|---|
| Car | 79.99 | 90.38 | 85.00 | 84.46 | 7.54 | 30 | 156 | 2337 | 942 |
| Pedestrian | 44.77 | 52.48 | 64.50 | 35.40 | 25.77 | 152 | 1133 | 8325 | 2525 |
You can find detailed results on KITTI test set on this website.
Multi-object tracking(on KITTI val)
| Category | HOTA (%) | MOTA (%) | IDS | FP | FN |
|---|---|---|---|---|---|
| Car | 82.91 | 91.96 | 1 | 302 | 371 |
| Pedestrian | 50.99 | 64.75 | 70 | 2240 | 1140 |
On KITTI, we use PointGNN as our detector.
License
CAMO-MOT is released under the MIT license.
Acknowledgement
In the detection part, many thanks to the following open-source projects:
- CenterPoint
- FocalConv
- BEVFusion
- We especially thank [email protected](FocalConv) for his help.
In the tracking part, many thanks to the following open-source projects:
Citation
If you find our paper useful for you, please consider cite us by:grin::
@misc{https://doi.org/10.48550/arxiv.2209.02540,
doi = {10.48550/ARXIV.2209.02540},
url = {https://arxiv.org/abs/2209.02540},
author = {Wang, Li and Zhang, Xinyu and Qin, Wenyuan and Li, Xiaoyu and Yang, Lei and Li, Zhiwei and Zhu, Lei and Wang, Hong and Li, Jun and Liu, Huaping},
keywords = {Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences, FOS: Computer and information sciences},
title = {CAMO-MOT: Combined Appearance-Motion Optimization for 3D Multi-Object Tracking with Camera-LiDAR Fusion},
publisher = {arXiv},
year = {2022},
copyright = {arXiv.org perpetual, non-exclusive license}
}