A collection of tools used to track and error-correct image bounding boxes generated by YOLOv4 algorithm, calculate camera-radar plane conversion matrix, convert camera bounding box to radar bounding box.
A set of more 'relaxed' version of environment dependencies is supplied to help with comptability on different systems.
This repository is in active development and things might change/break before they are finalized.
- Perform bounding box tracking using
image_tracking
- Manually find common points in radar and camera dataset (using the tracked boudning box data and
visualization.py
tool) - Calculate camera-radar conversion matrix using the tools found in
cr_transform
(requires MATLAB) - Export the transformation matrix to
radar_bbox_from_img
- Use
imag_to_radar_2.py
to convert camera bbox to radar bbox with 2D matched filter (In-progress) - Output radar label (Upcoming)