I am planning to build an app which will count the number of sqauts from videos. Assuming that the user and camera do not move, are there ways I can count the number of squats? Do such models to understand human activity and pose exist?
Yes, this is a rather straightforward problem that should be able to be solved using openCV or ANNs.
Check out this paper to get you started: https://arxiv.org/pdf/1701.08936.pdf