Facebook Announces “DeepFace” Photo Matching Technology

privacy_eyeFacebook has announced a new technology program called “DeepFace” that can determine whether two face photos belong to the same person. The site says the technology is correct 97.25 percent of the time, while a live human being scores just slightly higher at 97.53 percent.

You may not realize it, but Facebook is already sophisticated at recognizing you in your photos. They can suggest friends to tag in your uploaded photos by measure the distance between the photo subjects’ eyes and nose in profile pictures and tagged photos. However, Facebook wants DeepFace technology to go much further than that, eventually creating a “neural network” in the program. DeepFace would be able to create 3D models of users’ faces and rotate them virtually so they appear to always be looking at the camera, and then compare that information to that of an “average” face.

DeepFace is only a research project for now, which means it won’t be affecting you anytime soon. However, Facebook has repeatedly shown an interest in incorporating artificial intelligence, so it’s likely that the site will be able to closely analyze public photos and text one day sooner rather than later. The site has said in the past that it even wants to have artificial intelligence programs capable of gleaning mood and context from posts on the site one day.