bharath5673/3d_human_pose ? reverse-engineered prompt

Build me a simple Python project that can take a video file or webcam feed and estimate a person’s pose in real time, then draw the skeleton on top of the video and save the result to an outputs folder.

I want a few easy scripts, one using YOLO pose, one using YOLO plus MediaPipe landmarks, and one that shows a more 3D style pose view like the NVIDIA workflow. Keep it beginner friendly, so I can run each script without a lot of setup, and make it easy to swap in my own model files and input videos. Please include clear defaults, basic error handling if a model or video is missing, and make sure the output videos are actually written out.

If anything is unclear from the existing code or model docs, look up the current docs online and use the simplest working approach.