Deep learning network to detect activity in videos

Hi all, I'm currently working on a project to create a model that detects the type of activity two people in a video are doing. For example, let's consider a data set consisting of videos where the two people are boxing, wrestling, jumping, and even just talking with each other. How can I create a model to distinguish these activities from each other.

In particular, what kind of features should be extracted from the videos ? Should I use a …

deep learning deeplearning learning network videos