Cartoonizer is based on a research paper by University of Tokyo researchers Xinrui Wang and Jinze Yu. The tool leverages their open-source implementation to create a publicly-available demo of the method, using GPU (Graphics Processing Unit) servers for the video inference and CPUs (central processing units) for the images. It’s far from the first attempt to turn people into cartoons, but its videos are pretty impressive. Check out this cartoonized scene from Marvel’s Avengers: [Read: This AI spit outs endless new My Little Pony characters]
Becoming a cartoon
Being a narcissist intrepid reporter, I deigned to test the tool on my own countenance. When I fed it a headshot, Cartoonizer quickly spat back a rather haunted portrait — perfect for the brooding new look I plan to cultivate.
Unlike so many AIs, Cartoonizer also seems to work well on people of color:
Finally, to test Cartoonizer’s video converter, I fed the app a clip of my favorite athlete: LeBron James. Mahajan and Pandkar plan to open source the code and write an article on the architecture within the next few days. In the meantime, you can try the tool out yourself at the Cartoonizer website.
