Recently, Google built a computer that taught itself to recognize cat videos. That’s pretty neat. But what constitutes a cat? What are the basic elements that impart cattiness into an object?
Here’s a picture I drew of a cat face. Most people would look at that and understand it represents a cat.
Now, this drawing consists of 6 elements: head, ears, eyes, nose, mouth, whiskers.
Six elements can be combined in 64 different ways (2^6=64). Each element could be assigned a digit in a 6-digit binary number, so just the head might be 100000, and head & eyes might be 101000, and so on. Then you could count from 000000 (0) to 111111 (63) to get all possible combinations.
Here, I’ll draw them out for you…
Good thing I had spare cards.
Let’s carefully arrange them in order of least catty to most catty.
…or possibly just shuffle them around a bit, drop some on the floor and then give up.
The best I came up with was that the 6 drawings with one missing element were the next-most catlike. I’d say the least-important elements are the nose and mouth, and the most important is the ears. None of the drawings without ears looked much like a cat. And, weirdly, the ones without the head element looked like Yorkshire Terriers.
OK, sure, not a very scientific study and the sample size (1) is too small to get any useful statistics. But there’s definitely something interesting going on there. I think someone needs to expand this into a real study with proper funding and a cool title like “The Effects Of Element Masking On Facial Recognition In Graphical Representations Of Felines” or something. Kickstarter anyone?