AlphaPose provide three different tracking methods for now, you can try different method to see which one is better for you.
Currently the best performance tracking model. Paper coming soon.
Download human reid model and place it into AlphaPose/trackers/weights/
.
Then simply run alphapose with additional flag --pose_track
You can try different person reid model by modifing cfg.arch
and cfg.loadmodel
in ./trackers/tracker_cfg.py
.
If you want to train your own reid model, please refer to this project
./scripts/inference.sh ${CONFIG} ${CHECKPOINT} ${VIDEO_NAME} ${OUTPUT_DIR}, --pose_track
- [] Evaluation Tools for PoseTrack
- [] More Models
- [] Training code for PoseTrack Dataset
Use a human detecter with tracking module (JDE). Please refer to detector/tracker/
Download detector JDE-1088x608 and place it under AlphaPose/detector/tracker/data/
Enable tracking by setting the detector as tracker: --detector tracker
./scripts/inference.sh ${CONFIG} ${CHECKPOINT} ${VIDEO_NAME} ${OUTPUT_DIR}, --detector tracker
This tracker is based on our BMVC 2018 paper PoseFlow, for more info please refer to PoseFlow/README.md
Simply run alphapose with additional flag --pose_flow