Well, that’s one use of AI I can get behind! A group of researchers at Microsoft have unveiled their artificial intelligence powered application that does some unique little artistic magic.
Basically, the app takes ordinary photographs of people and turns them into hilarious caricatures.
As reported, the researchers are Stanford graduate Kaidi Cao, Lu Yuan from Microsoft AI Perception and Mixed Reality and Jing Liao from the City University of Hong Kong. They are expected to unveil the project next month at SIGGRAPH Asia in Tokyo.
And while they admit the AI is not perfect, but around a quarter of the images generated by it can be mistaken for actual human created art.
Observe:
Not bad.
More than a thousand different photographs were used to train the AI for this.
The work employs a type of AI algorithm known as generative adversarial networks (GAN), and utilizes the favored deep learning technique for creating deep fakes. It is used in manipulating real video and images to create convincing fakes.
Instead of creating deep fakes, however, the researchers used GANs to create a cartoon generator capable of outputting caricatures like a human artist.
Or close enough, what have you.
The researchers specifically focused on shape exaggeration and appearance stylization, as these two are the key aspect to drawing a caricature that is still recognizable when wrapped.
Take a closer look at the research here, or give this PDF file a read if you want in on the technical nitty gritty of this AI implementation. While it’s not yet clear what the practical implementation of a technology like this can be, it still is a rather impressive step forward for artificial intelligence.
And, at the very least, gives us an excuse to put up the caricature image above.
On Turkey Day, no less.
