The Future of Computer Vision: How Two Rubyists Are Changing The World
summary
Two implementations for scanning and identifying Magic: The Gathering cards are described: a fixed dispenser built from LEGO Mindstorms that feeds cards to a mounted camera, and a cheap lightbox plus webcam workflow that uses OpenCV to detect, warp and crop cards. Recognition uses perceptual hash fingerprints stored by multiverse id and compared by Hamming distance (threshold tuned to ≤6) with human confirmation to grow the corpus. Both demos work; the LEGO rig is fragile and expensive, the lightbox is cheaper but requires more image processing and manual feeding.
AI-written summary · about
chapters (16)
- 0:00Future of computer vision
- 1:19Speakers and magic card problem
- 5:30Solutions overview
- 6:53Card identification with AI
- 8:37Perceptual hashing for images
- 12:49Divergent implementations
- 13:09Urza Lego robot for scanning
- 16:41Mining Magic card data
- 18:05Three card identification methods
- 20:51Freestyle system with OpenCV
- 23:18Eight steps to extract cards
- 28:55Live demo freestyle system
- 32:05Future work and ML goals
- 34:12Why systems couldn't merge
- 35:56Live demo Urza robot
- 39:24Q&A
related talks
29mA Magical Gathering
43mOh, Oh, Oh, It's Magic!
26mIs This Your Card? Computer Vision for Playing Card Recognition
40mComputer Vision Using Ruby and libJIT
32mMagic of Computer Vision Unraveled
26mDetecting and classifying object images using ruby
34mMaking a Robot Eye with JS or Magic of Computer Vision Unraveled
30mIntroduction to Computer Vision with OpenCV