"Lights, camera, AI!" - Automated sports videography
summary
Builds a Raspberry Pi–mounted camera plus servo that points at play, runs a compressed machine‑vision pipeline (frame differencing then column summing) and an MLPRegressor with ~20 inputs → 33 hidden → 1 output, yielding R² ≈ 0.8. Labels ~6000 images collected at ~3 fps, gets predictions at ~2 fps, uses Python multiprocessing to smooth servo movement, and demos a working prototype that mostly follows rugby action though it sometimes lags.
AI-written summary · about
chapters (14)
- 0:00Automated sports videography
- 1:05Physical computing tools
- 4:12Rugby filming problem
- 6:17Automated filming robot
- 8:48Labeled data for ML
- 11:55GUI for rugby footage
- 13:40Feature engineering
- 18:35Training neural network
- 20:21Model on Raspberry Pi
- 22:49Smoothing camera movement
- 25:19Startup automation
- 26:43Live demo and testing
- 28:31Future possibilities
- 29:59Q&A on model performance
related talks
37mUnstructured Data Processing with a Raspberry Pi AI Kit and Python
51mVideo Analytics for Football games
27mAnalysis of sports scenes using computer vision models
42mDeep Learning for Automatic Basketball Video Production
35mFootball video analysis using Deep Learning
44mAdventures in Puppy Training with A.I. and a Raspberry Pi
28mChasing birds with code
21mLeveraging Python for Real-Time Image processing