Tool was created by feeding a computer pictures from Flickr, and letting it learn what people look like

Facebook has developed an algorithm that can recognise who people are even if they cover up their faces.

The sophisticated technology can spot people based just on what they’re wearing, or the shape of their body, according to the New Scientist. It could eventually be put to work to identify the people in pictures uploaded to Facebook.

The site already has a range of sophisticated technology that can spot people’s faces, tagging them with their name and creating special collections of images from certain situations. But the team behind the technology looked to find out whether the robots that do so could, like humans, recognise someone without actually seeing their face.

To do so, researchers fed the neural network 40,000 public photos from Flickr. That created an algorithm that let it recognise people with 83 per cent accuracy.

Facial recognition technologies have been a point of fierce debate — and the features that use it are turned off in Europe, because of regulators’ concern about the technology. For instance, Moments — a tool developed to sort pictures of the same event into one collection for easy viewing, is turned off in the EU.