Rotated-Fisheye Tracker is a simple motion tracker that transforms rotated bounding box detections into 2d Gaussians distributions and then uses distribution distances to associate which replicates the IoU association behaviour.
Step1. Install RF-Tracker.
git clone https://github.com/Skyteens/RF-Tracker.git
cd RF-Tracker
conda create -n <ENV_NAME> python=3.8
conda activate
pip install -r requirements.txt
Run RF-Tracker:
- prepare a folder of each video frame and its rotated bounding box detections
cd <RF-Tracker_HOME>
python3 demo.py
- For different association methods
python3 demo.py --match [gwd,bd,kld]
A large part of the code is borrowed from ByteTrack, MMRotate, and RAPiD. Many thanks for their wonderful works.