Hi,
I am working on a similar project aiming to understand individual/group activities from surveillance videos. I read the thesis and found it quite relevant to what I need. Would you mind releasing more trained models (like the one you shared for orientation estimation) for other tasks? Thank you!