Mixformer2

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

Usage

Pre-requsite

docker

nvidia docker

OpenCV

cmake (optional)

1. Build the image for tracking server

sudo docker build --network=host -t tracking_server .

2. Run the container

sudo docker run --gpus all -p 8002:8002 -it --rm tracking_server:latest

3A. Run python client

python python_client.py /home/nicky/cutted_football.mp4

3B. Run C++ client

mkdir build
cd build
cmake .. 
make 
./cpp_client /home/nicky/cutted_football.mp4