Slightly modified version of the pytracking repository. Make sure to clone this repository with the --recurse_submodules flag and initialize the submodules:
git clone --recurse-submodules <link-to-repo>
cd pytracking
git submodule update --init --recursive
In order to use use any of the trackers, it is necessary to build the pytracking docker container. Follow the steps described in the README.md and find the mounted project repository in the docker root directory, once it is running.
To finalize the setup of pytracking make sure to run the install script. The install script downloads the wished network models for the trackers. Per default, the networks associated with the RTS tracker are downloaded. The install script also sets up the environment by running the create_default_local_file() in pytracking and lts which create local.py files which describe where the pretrained networks are stored for example.
./install.shAfter running the install script, make sure that the paths in the local.py files for pytracking and ltr are set correctly, according to your environment.
The nice thing about the RTS tracker is that one can interactively draw a bounding box around the object one wants to track while the tracker is running. In order to do so, navigate to the pytracking directory and run run_python.py.
cd pytracking
python3 run_video.py rts rts50 <path-to-video> --debug=<debug-level>
To test if everything is working properly, navigate to the pytracking directory and execute the run_video.py file with a demo video of a drone flying. Click at one corner of the drone and see a bounding box apppear in blue. Click a second time diagonally from the initial click to fully enclose the drone. Hit space to run the tracker frame-by-frame.
cd pytracking
python3 run_video.py rts rts50 experiments/demo.mp4