Ask Your Question

Best approach for identifying Set cards

asked 2014-12-27 18:14:01 -0600

I have some basic experience with OpenCV after reading a few books. I want to write a program that identifies cards from the game Set from an image like this (and eventually from live video):

Set cards

I'm wondering what the best approach would be.

  • Should I train nine objects (three different shapes with three different shadings) and find those in the image?
  • Should I use edge detection and try to identify the shapes that way?
  • Is there some other approach I should use?
edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted

answered 2014-12-29 07:20:52 -0600

Guanta gravatar image

First of all: cool game, I liked it a lot, though I am not pretty good at it ;)

For your questions: It's really difficult to tell which approach will be the best one. I wouldn't go with a Cascade-Classifier in that case since I guess you'd a lot of mismatches then. So I'd start with a simpler approach:

  • First locate each card in the image, maybe the following link helps you with that:
  • Then you need to classify each detected card, here I would start with simple features: for each color channel r,g,b compute mean and standard deviation and compare them with all others (i.e. you need to compute those in advance for all the cards). Pick the one with the nearest Euclidian distance. If that won't give you good enough results, then add e.g. moment-based features which would correlate to the shapes you have.

Good luck!

edit flag offensive delete link more


Thanks for the suggestions. I was traveling over the holidays and unable the start work on it. But I'm going to try to make some progress over the next couple weeks.

SSteve gravatar imageSSteve ( 2015-01-06 18:06:10 -0600 )edit

Question Tools

1 follower


Asked: 2014-12-27 18:14:01 -0600

Seen: 436 times

Last updated: Dec 29 '14