This program takes a video input from webcam or video file predict poses using torso, face and hand detection. This program helps to identify 9 different human body poses with a frame rate of upto 15 frames per second on a i-5 4110 CPU. It succesfully predict 9 different poses with an acuracy of more then 80%. Required softwares for the program to run:
- Opencv (2.4.13)
- Numpy (latest)
- Sklearn (optional if you want to use the training model that has been uploaded)
- Pickle
Place all the files in a folder. After installing these libraries/sofwares run detect_poses.py Make appropiate changes in VideoCapture function (which type of input you want '0' for webcam and 'filename' for recorded video feed).