This repository is based on MCG-NJU/MixFormerV2.
As there is no any inference framework that support real time online learning, therefore the main goal here is to build a tracking as a service with python client and C++ client.
The code for server is in /mixformerv2/tracking/app.py
docker
nvidia docker
OpenCV
cmake (optional)
sudo docker build --network=host -t tracking_server .
sudo docker run --gpus all -p 8002:8002 -it --rm tracking_server:latest
python python_client.py /home/nicky/cutted_football.mp4
mkdir build
cd build
cmake ..
make
./cpp_client /home/nicky/cutted_football.mp4