There are a lot of reasons why you wouldn’t want your drawings to come to life. If you need another, check out Christopher Hesse’s edges2cats. The project, which was created with Google’s open-source machine-learning project called Tensorflow, is an image-to-image translation that produces “real” looking cats from your doodles—yet the results aren’t so much cute and cuddly as they are haphazard monstrosities featuring random tufts of fur and distorted eyes.
The algorithm was trained with over 2000 photographs of cats, which gives it the ability to see lines and guess whether they’re supposed to be eyes, tails, or limbs. From there, it appears to grab the clone Photoshop tool and go to nightmare town.
“Some of the pictures look especially creepy, I think because it’s easier to notice when an animal looks wrong, especially around the eyes,” Hesse writes on the site. “The auto-detected edges are not very good and in many cases didn’t detect the cat’s eyes, making it a bit worse for training the image translation model.”
If inanimate objects are more your speed, Hesse also has programs that reproduce buildings, bags, and shoes, which produce less terrifying results. But since the edges2cats program yields the most visceral reactions, the mental_floss team made a few examples of how AI can go horribly (and still wonderfully?) wrong:
[h/t The Verge]
February 22, 2017 – 3:30pm