Mask R-CNN and Keypoint R-CNN wrapper + BYOL and SimSiam wrapper for Multiple Object Tracking with PyTorch.
rcnnpose-pytorch-tracker/
ㄴ .idea/
ㄴ examples/
ㄴ media/
ㄴ KalmanFilter.py -- additional method
ㄴ eval_MOT17.py -- evaluation entire MOT Challenge datasets
ㄴ tracker.py -- tracker main
ㄴ tracker_utils.py -- tracker utils.
ㄴ video_demo.py -- demo(video)
ㄴ image_demo.py -- demo(image)
ㄴ rcnnpose
ㄴ __init__.py
ㄴ estimator.py
ㄴ utils.py
ㄴ simsiam
ㄴ models/
ㄴ simsiam_standalone.py
ㄴ trackeval
- Install PyTorch
>= 1.2
with torchvision>= 0.4
- Download ckpt and unzip at ./simsiam/ckpt/
- CUDA
>= 10.0
, Cudnn>= 7.6.5
(test env) - TrackEval -> officail python evaluation code & data
git clone https://github.com/jimilee/rcnnpose-pytorch-tracker.git
pip install -r requirements.txt
cd examples
python image_demo.py
python video_demo.py
python tracker_demo.py