Turn any live video stream or locally stored video into a dataset of interesting samples for ML training, or any other type of analysis.
Why do you think that https://github.com/automaticdai/rpi-object-detection is a good alternative to automatic-video-processing