How can I perform Human Activity Detection from Live Video?

I am working on a project with MATLAB to detect human actions from live video using deep learning techniques. I am a beginner and I would like to learn the possible methods, so that it will be helpful for me to achieve my target. I have been trying this with CNN and YOLOv2 but I cannot achieve what I actually wanted. Could anyone suggest me some ideas in step by step process from Collection of Dataset to Training the model and Detection?