Facebook's new experimental algorithm can identify you even from unclear pictures

New facial recognition software can identify people with 83% accuracy using hairdos, clothing, body shape etc

Have you ever wondered how Facebook can automatically tag your friends in the pictures you post onto the social network? A secret concoction of code and powerful facial recognition has now enabled the social network to track people, even when they may not want to be tagged.

Yann LeCun, head of artificial intelligence at Facebook, has brewed a new algorithm and is currently experimenting with it. It can recognise people in photographs even when their faces are not quite visible or obstructed. That is because this artificial algorithm looks for distinct features like one’s hairdo, clothing, and body shape, posing styles in addition to just the face and directly recognisable features.

Modern face-recognition algorithms, which are already prevalent in social networks, shops and even churches, have made developers like LeCun question if something like that can be replicated for the social media network.

"There are a lot of cues we use. People have characteristic aspects, even if you look at them from the back," LeCun explains. "For example, you can recognise Mark Zuckerberg very easily, because he always wears a gray T-shirt."

To make it a success, the research team at Facebook ran a test on some 40,000 public photos – of people with their faces clearly visible – from photo sharing site Flickr including photos in which they were not looking into the camera or were actually turned away.

After running them thorough a sophisticated neural network, they found out that the final algorithm was able to recognise people in photos with around 83 per cent accuracy.

The results were then presented earlier this month at the Computer Vision and Pattern Recognition conference in Boston, Massachusetts.

The algorithm can help power photo apps like Facebook's Moments, released a few days ago

However, the social media network will have to address certain ethical issues surrounding such an app. The algorithm’s ability to recognise people who are not looking into the camera, or worse are trying to hide their faces deliberately, will raise acute privacy issues. Something Facebook is quite used to by now.

The manifestation of this issue at hand was seen during the talks over rules governing facial recognition which collapsed after privacy advocates and industry groups did not agree. The talk was held in Washington, DC organised by the US Department of Commerce National Telecommunications and Information Administration.

“If, even when you hide your face, you can be successfully linked to your identify, that will certainly concern people," says Ralph Gross at Carnegie Mellon University in Pittsburgh, Pennsylvania, who says the algorithm is impressive. "Now is a time when it's important to discuss these questions.”

This article originally appeared on New Scientist