You know when you find something and think, “Shit. There goes the weekend”. Recently, that happened with Deepart.io. A site I just found that lets you upload two images and have neural nets “paint” one image with elements of the other. It takes some tinkering to find images that work well. But I soon got the AI doing some pretty amazing things. Below is an image I created with a pic of a friend doing yoga and some street art.
I also found that choosing a good “target” image is key. For example, I found an image of a bike that resulted in predictably cool results.
If you’re curious to see where the rest of my weekend went, check out my Deepart Flickr album.
I have to say, finding this project was a little bittersweet. The aesthetic it excels at is very similar to the aesthetic I was going for with my Fractal Mosaics project, which never quite took off. It has that same fractal-like quality, where objects in images seamlessly merge with another. Similar to some of my favorite painters. At the same time, it was a thrill being able to explore that aesthetic “at scale” (i.e. much faster than my clunky Fractal Mosaics code allowed). Hats off, Deepart.io. This is truly revolutionary. Can’t wait to see what’s next.