Making a Robot Eye with JS or Magic of Computer Vision Unraveled
summary
Builds an AR marker detector in plain JavaScript using canvas and video input, reads getImageData RGBA array, converts to luma grayscale + threshold to pull bright regions, runs blob labeling and edge tracing to find connected components and corners, computes a homography to warp the distorted patch and does a pixel‑by‑pixel match against a reference (requires ~100px marker). Uses Web Workers to keep UI responsive; demo works in the browser though code is simple and not performance‑optimized.
AI-written summary · about
chapters (9)
- 0:00Intro
- 0:55Personal
- 3:15Project
- 6:03Demo
- 8:54Canvas
- 11:22Image Data
- 17:06Grayscale and Threshold
- 21:32Labeling and Adding
- 29:47Performance
related talks
32mMagic of Computer Vision Unraveled
26mSelf-Portraits in JavaScript
32mSelf-Portraits in JavaScript
49mReal Time Face Detection through a Webcam Using AI
51mRealTime Exoskeleton-Mapping and Body Segmentation on the Browser with Tensorflow.js
56mTensorFlow.js: Make a smart webcam in JS with a pre-trained ML model
41mThe Future of Computer Vision: How Two Rubyists Are Changing The World
33mComputer Vision, Not Just for Breakfast Anymore