Xerox Researchers Develop Algorithm that Can Detect Pretty Images

If you’re anything like most digital camera owners, you are probably taking thousands of photos every year. Sorting through those to find the best images, though, can become a bit of a hassle. What if there was an algorithm that could look at your photos and automatically pick the best ones out for you? Developers at Xerox’s research center in Europe are working on just this kind of algorithm and just released a first demo of what this would look like.

For the most part, Xerox argues, the current crop of image classification algorithms have focused on recognizing and tagging certain objects in images. Finding images that are pleasing to the eye, though, necessitates a very different approach. According to Xerox, “this new prototype demo tackles the difficult task of trying to learn what makes an image special and unique. For example, good beach photos are often characterized by silky waves obtained by a long camera exposure.”


You can play with a demo of the new algorithm here, though you sadly can’t import your own photos yet. Xerox told Technology Review, though, that it plans to beta test this technology with some of its partners next year.

For now, the technology seems to have a bit of a fetish for black-and-white images and of course it doesn’t exactly prefer those family snapshots that may not be highly aesthetic but capture some of the more important moments in life. In return, though, this algorithm could help you to finally sort those 3,000 pictures you took during your last vacation…