This Python program detects individual people from a video frame using YOLO, crops the frame, and passed the frame through Mediapipe pose detection to get the coordinated of the individual people.
The program runs very slow currently running at around 1.5-2 fps on CPU. Could use some GPU acceleration.
YOLOV3.weights file excluded due to large size. The video files have to be downloaded separately and have been excluded for privacy and copyright issues.