Facebook’s extremely accurate facial recognition technology, which enables the program to automatically tag faces in pictures uploaded to the website, could be getting an upgrade. The social networking giant has developed experimental technology that can work out who you are even if it can’t see your face, according to New Scientist.
The tech looks for other possible identifiers — regularly-worn clothing, a distinctive posture, or a signature haircut — and uses these clues to guess the person in the photo.
The technology is being developed at Facebook’s artificial intelligence lab, the New Scientist says. In their tests, the team’s software was able to recognise individuals with 83% accuracy, even if the faces were obscured.
The tests were based on a sample of 40,000 photos taken from Flickr. This is far smaller than the 250 billion photos the social network has sitting on its servers as of 2013.
“There are a lot of cues we use,” Facebook’s head of AI told New Scientist. “People have characteristic aspects, even if you look at them from the back … For example, you can recognise Mark Zuckerberg very easily, because he always wears a grey T-shirt.”
Facebook is currently doubling down on its facial recognition tech. It recently launched Moments, an app that scans a user’s photo roll on their phone, detects who are in the photos, and offers to send them to them. This provides users with an easy way to share lots of photos at once — and gives Facebook access to large quantities of new image data in the process.
However, there have been hiccups. Facebook’s tag suggestion service that detects who might be in a photo and offers to tag them isn’t operational in Europe because of privacy concerns by regulators and watchdogs. As such, Moments is not currently available on the continent, and the social network has no immediate plans to release it.
NOW WATCH: Thanks to Facebook’s virtual reality device, I’ve been to ‘The Matrix’ and all I want is to go back
Business Insider Emails & Alerts
Site highlights each day to your inbox.