In addition to its Cloak and Scribbler projects, Adobe also used its MAX 2017 conference to offer a sneak peek of a technology called Scene Stitch. It’s like Content-Aware Fill on steroids: instead of guessing the fill content with details from the photo, Scene Stitch uses AI and a database of images to find content to fill the hole.
“Scene Stitch isn’t just matching image types, it’s looking for what would fit the image well.”
One of the examples given in the demo is this photo that contains a road that needs to be removed:
Using standard Content-Aware Fill results in this not-so-ideal result that contains obvious cloned sections:
But run Scene Stitch, and the software will search through Adobe’s gigantic library of stock photos for images with sections that may fit well into the hole. What you get returned is a selection of options to choose from:
Many of them are much cleaner fills than the Content-Aware Fill (except the fills are from other photos):
Here are a couple more before-and-after examples:
“This is doing more than just filling the hole in the image,” says Adobe’s Brian Price in the demo. “This is really allowing you to reimagine the image, to try new things, to remix in new content.”
As with many of Adobe’s latest technologies, Scene Stitch is powered by Adobe Sensei, the company’s powerful artificial intelligence system.
No word yet on if or when we’ll be seeing Scene Stitch arrive in the real world as a feature photographers and photo editors can actually use.