<< return to Pixycam.com

Detecting playing cards


using the PixyCam 2.1, I’m trying to at the very least recognise playing cards; not necessarily the values or symbols (although this would be my end goal). I’m wondering if;

  • CC can be used on playing cards (e.g. to recognize the card back, or value).
  • shadows will always remain a problem for the pixycam, or if there’s something I can do about it
  • lower the amount of incorrect detected objects… I’m guessing this is hard since a card is basically a white rectangle. Lowering the range (slightly) also causes the cards to go undetected too soon sadly.
  • cards can also be easily recognised on a different surface than what I taught the camera on; a nice contrast of white cards and a blue surface works quite well, compared to cards on my beige wood table…

I’ve been reading the documentation, and other objects seem to work better… so fingers crossed I’m missing something helpful!

Any help would be much appreciated!

I think detecting playing cards might be possible, but recognizing the card face is probably beyond Pixy’s capabilities. Pixy can detect objects using color cues and it can detect some other things (lines). It sounds like you want to detect some letters and numbers. Pixy might be able to detect if it’s a king, queen or jack vs a numbered card. Sounds like a fun project :slight_smile:


1 Like

Thanks for your response!

Just thought I’d let you know that detecting red symbols on a white card is certainly possible;
And since the value of the card equals the amount of symbols detected, I’m now also able to work out the card’s value!

The spade & club symbols though are black, and I’m sure you already know; hard to detect… Unless you have any advice on this specific issue, I think I’ll keep the pixycam in storage for another project, and maybe use a basic camera to work out the pixels myself, since I’m only working with a flat surface.

Ah, I hadn’t considered! Counting the symbols is a really good way of determining the face of the card :slight_smile Our apologies about black though. We don’t detect black or white because these don’t tend to be good cues to detecting objects (they could be overexposed regions (lights) or they could be underexposed (shadows, etc)), but I see that your application could make good use of detecting black.
Thank you for describing your problem :slight_smile: