Data scientist and photographer Bhautik Joshi has been experimenting with computer-generated style transfers: feeding a neural network an example of art in a particular style and then asking it to redraw other images the same way. In this case, the model was one of Pablo Picasso’s 1955 series Les Femmes d’Alger, and the images to redraw were frames from Stanley Kubrick’s masterpiece 2001: A Space Odyssey. The astonishing result, which you can see above, is a rotoscope effect beyond the HAL 9000’s wildest dreams.
According to Joshi’s Tumblr, his style transfers are based on the research of Leon A. Gatys, Alexander S. Ecker, and Matthias Bethge. He’s not the first person to create this sort of animation: Artist and programmer Gene Kogan has done similar work. Here’s Kogans animation of Alice in Wonderland in the style of 17 different paintings:
The real test, of course, will be rendering films about painters in their own style. Frida’s Frida! Andrei Rublev’s Andrei Rublev! The possibilities are endless—just don’t ask the computers to animate Thomas Kinkade’s Christmas Cottage or we’ll have a Skynet-type situation on our hands.
