Visualization Tools

March 27, 2023 ยท View on GitHub

Want to know how we made the beautiful demos? Our programs are in ./tools/video_demo/.

1. Multi-camera Tracking Video

  • INFO_FILE_PATH: path to the infos file.
  • TRACKING_RESULT: path to your tracking result file in .json format.
  • SHOW_DIR: directory to save the images and videos.

The command is:

python tools/video_demo/cam_demo.py --data_infos_path INFO_FILE_PATH --result TRACKING_RESULT --show-dir SHOW_DIR

For example, if you want to visualize the results on the validation set of mini-split for results.json, please run the following commands:

python tools/video_demo/cam_demo.py --data_infos_path ./data/nuscenes/tracking_forecasting-mini_infos_val.pkl --result results.json --show-dir ./work_dirs/visualizations/

2. Bird's-eye-view (BEV) Video

By projecting the 3D bounding boxes to the BEV space, we can assess the quality of boxes more easily. Please remember to install SimpleTrack as "environment set" to use our BEV visualization program.

  • CONFIG_PATH: configuration path.
  • TRACKING_RESULT: path to your tracking result file in .json format.
  • SHOW_DIR: directory to save the images and videos.

The command is:

python tools/video_demo/bev.py CONFIG_PATH --result TRACKING_RESULT --show-dir SHOW_DIR

For example, if you want to visualize the results on the validation set of mini-split for results.json, I will run the following commands:

python tools/video_demo/bev.py ./projects/configs/tracking/petr/f3_q500_800x320.py --result results.json --show-dir ./work_dirs/visualizations/

3. Notes for Developers

  • Point clouds. In BEV visualization, we overlay the point clouds for visual clarity.
  • Color of boxes. We assign a fixed color for each ID as color = COLOR_MAP[COLOR_KEYS[track_id % len(COLOR_KEYS)]] so that the consistency of colors reflect the quality of tracking.
  • Motion prediction. In BEV visualization, we also visualize the predicted trajectories.