A look at how the DeepFace software recognizes and verifies the human face in photos.
When given two unfamiliar photos of faces, humans can figure out if they're the same person or not 97.53 percent of the time, according to the MIT Technology Review. Now with Facebook's new software, computers can find it 97.25 percent of the time, too.
This new artificial intelligence, also dubbed DeepFace after the term deep learning, has been improved over time to approximate human performance, according to Yaniv Taigman, a member of the Facebook AI team. "You normally don't see that sort of improvement," he told the Review. The new deep learning uses "simulated neurons" to recognize patterns in lots of data. Essentially the software isn't recognizing faces but verifying facial structure.
All this just to figure out who to tag in a new photo? Luckily for most of us, it's still in the developing stages (despite the paper
which will be presented at the IEEE Conference on Computer Vision and Pattern Recognition in June.) “We are publishing our results to get feedback from the research community,” Taigman told the Review. Taigman developed DeepFace along with Facebook colleagues Ming Yang, arc’Aurelio Ranzato and Tel Aviv University professor Lior Wolf.