Magic of Computer Vision Unraveled
summary
Concrete recipes and tradeoffs: uses HTML5 canvas getImageData to read RGBA pixels, applies per-pixel LUT filters, convolutions (box and Gaussian blur), thresholding, connected-component labeling and border tracing to get area/circumference, fingertip detection by angle and center-of-mass, and circularity shape factor for shape ID. Claims JavaScript stalls UI at large sizes (400×300 ok, 2000×1000 kills it) and moves heavy work to webworker or GPU via WebGL; ships a helper library graphic.js and points to Book of Shaders.
AI-written summary · about
chapters (8)
- 0:00Non-verbal communication intro
- 4:22Pixels to social constructs
- 10:23Color filters and changes
- 14:18Object identification
- 20:21Hand gesture detection
- 24:37Browser image processing pitfalls
- 27:24Web Workers offloading
- 30:12Further learning
related talks
36mMaking a Robot Eye with JavaScript
34mMaking a Robot Eye with JS or Magic of Computer Vision Unraveled
16mImage Processing in JavaScript
26mHow to Cheat at Life with JS, CV, and AR
59mYou can't do that with JavaScript!
26mRecreate Masterpieces of Modern Art with JavaScript!
37mRendering performance inside out
59mYou can't do that with JavaScript!