Awesome
3D MOT on Waymo dataset
This project is developed for online 3D multi-object tracking on Waymo dataset. It is build upon CasA
detector and CA-based KF tracker. The tracking code
is from here; detector is from here; visualization code is from
here.
Waymo Results
Vehicle/Pedestrian/Cyclist online tracking results, evaluated by MOTA/L2 metrics of Waymo benchmark.
- 2022/9/23 Update CasTrack-waymo, currently ranks third on Waymo tracking benchmark :fire:!
| set | Vehicle_L2 | Pedestrian_L2 | Cyclist_L2 | ALL_NS_L2 |
|---|---|---|---|---|
| val set | 59.30 | 62.78 | 61.72 | 61.27 |
| test set | 63.66 | 64.79 | 59.34 | 62.60 |
Prepare data
You can download the CasA detections from here (Please register with waymo before downloading, as the Waymo open dataset is under strict non-commercial license). To run this code, you need prepare the waymo dataset infos generated by CasA.
Quick start
- Please modify the info path, detections path in the yaml file to your own path.
- Then run
python3 waymo_tracking.py config/config.yaml - The results are automatically saved to
save_pathin yaml file, and evaluated by Waymo metrics (the results are lower than the one calculated by Waymo benchmark). You can create a submission file to evaluate the performance on Waymo benchmark by runningpython3 create_submission/create_submission.py