Tracking road users using off-the-shelf deep learning models
Thanks to the nice guys at Ultralytics, deep-learning computer vision applications using state-of-the art detectors are now much more accessible (and does not require one to use years old specific versions of scientific libraries).
At Polytechnique, we have developed a simple wrapper to track road users in video data, as a plug-in replacement for our old feature-based tracker. The tool is a Python script named dltrack.py (code repository). The documentation has been revamped to introduce it (and still needs a lot of work).
Take a look and report any issue (and contribute!).