Privacy-conscious people may want to start removing their photos from Facebook: a new experimental algorithm coming right from the social media giant’s artificial-intelligence labs is able to identify people with 83% accuracy even if their faces aren’t visible, reports News Scientist.
The algorithm was presented earlier this month at the Computer Vision and Pattern Recognition conference in Boston, Massachusetts. The catch is that people have other unique characteristics than their faces. Think hairdo, clothing, body shape, and pose.
“There are a lot of cues we use. People have characteristic aspects, even if you look at them from the back,” LeCun says. “For example, you can recognise Mark Zuckerberg very easily, because he always wears a gray T-shirt.”
To test the algorithm, the guys from Facebook’s AI labs looked at 40,000 public photos from Flickr: some of the people with their faces clearly visible and some with their faces hidden. Of course, they used them to find out how efficient the algorithm was.
The results were astonishing: the algorithm was able to identify the person with 83% accuracy even with his or her face hidden.
Now, if we look forward in time, the final version of this algorithm could be a power feature of Facebook’s Moments, which goes through the photos stored on your device, sorts them into separate events, and tags your Facebook friends it recognizes.
You know such a tool is double-edged: it would be great for privacy-conscious users to immediately spot any photos about them posted on Facebook, for example, and take the necessary steps, but it could also be used to identify someone even if he or she looks away from the camera.