The feature is being rolled out globally, although currently it is only available to some users. The new technology, called “DeepFace,” was developed for Facebook by an Israeli company called face.com, acquired by the social networking giant back in 2012. According to the company, it is able to recognize a human face in a new photo by comparing it with a previously uploaded photo with 97.25% accuracy. That accuracy rate is comparable with good old-fashioned “human looking at pictures” technology.
DeepFace (which I find to be a rather unfortunate name) works by constructing a three-dimensional model of a face based on a photo, then “rotating” it so the software can compare it to other photos from a variety of angles. The company tested the technology by having it match tagged photos with a database of over four million images, representing over 4,000 different people. Previous studies have shown that Facebook photos where users have been tagged tend to generate more comments and “likes” than photos without tags.
Since facial recognition still carries with it the inevitably creepy associations, Facebook is wisely enabling users who don’t want to be identified by the technology to opt out. If you don’t want Facebook to automatically suggest your name for tagging when people upload photos that resemble you, you can just go to the your Privacy Settings and change the parameters for photo tagging suggestions.
Israeli companies have long been leaders in facial recognition software, largely as a result of early investment by the Israeli military for security applications. Back in 2013, one widely discussed military system could supposedly scan a face and compare it to a database of known faces at the rate of 36 million faces per second -- but the technology is evolving rapidly and this system may already be out of date.