Adityabhatt1002/Football-AI-Project — reverse-engineered prompt

Build me a football match analysis project that can take a match video and turn it into tactical insights using computer vision.

I want it to detect the ball, players, goalkeepers, and referees, track each player with consistent IDs, separate players into two teams from their kits, and place everything onto a top down football pitch view. It should also detect pitch keypoints from the broadcast camera angle and use that to map player positions into real pitch coordinates.

Please include a clear notebook or main pipeline that runs end to end, plus training scripts for the player detector and pitch keypoint detector. The final output should show useful visuals like a FIFA style radar view, a Voronoi control map, and saved demo images or videos in an assets folder.

Use modern computer vision methods like YOLO style detection, ByteTrack style tracking, embeddings for team clustering, and homography for the pitch mapping. Keep the code organized and make it easy to install from requirements.txt.