"Sample" project that uses ScaledYOLOv4 and DeepSORT as packages on a video file, with option to save out human crops and inference video.
- Build Dockerfile and
./run_docker.sh
- ScaledYOLOv4 and DeepSORT are downloaded in the docker image by default, but if you need to edit the codebase, clone the repos on your machine (and download the weights). Inside docker, run
./setup_packages.sh
(decide if you want to install the packages as editable)
- Change the
cam_name_func
inmain.py
so that all videos invid_list
end up with a unique name. - TADAA!