Is This Your Card? Computer Vision for Playing Card Recognition
summary
Live webcam feed processed with OpenCV, numpy and imutils converts frames to grayscale, applies bilateral blur, then Canny edge detection and dilation to thicken contours. Contours are filtered by area and an aspect ratio 1.43 check to find the largest rectangular card, extracting a bounding box. Recognition uses thresholded face/number templates compared by pixel-wise template subtraction against a small trained set (one deck). No deep learning, runs in real time with occasional false positives but reliably finds and identifies a single card.
AI-written summary · about
chapters (13)
- 0:00Intro
- 0:30Disclaimer
- 1:18Premise
- 2:38Prestige
- 3:05Magic
- 3:19The real magic
- 4:09Computer vision
- 5:18Localization
- 7:42Normalization
- 13:20Dilation
- 15:50Find Contours
- 17:50Find Rectangles
- 19:15Aspect Ratio
related talks
14mIdentifying Pokémon Cards
41mThe Future of Computer Vision: How Two Rubyists Are Changing The World
43mOh, Oh, Oh, It's Magic!
30mIntroduction to Computer Vision with OpenCV
21mLeveraging Python for Real-Time Image processing
06mProviding Better Feedback in Real-time Object Detection Apps
22mShape Recognition with Computer Vision
57mFoundations of Facial Recognition