Hello,
Thanks for sharing your research and the training code!
Currently, I see that the repository mainly focuses on the training process. Would it be possible to provide a demo or inference script that allows us to run the trained model on custom video files? Thank you for your time and help!