Facebook has learned to recognize people without a face

Facebook facial recognition technology have reached such a level that they are able to identify a person even if the person is absolutely not visible (for example, from the back). This was made possible thanks to recent advances computer vision systems, пишет New Scientist.

Experimental neural network developed at the AI ​​lab, train for additional features such as hairstyle, hair color, clothing, shape and posture.

Conventional face recognition system are already working well enough, they caught up with the man, and some even walked a man. The Facebook decided to go ahead and try to teach the program to recognize people, even if the face is difficult to see or do not see, because people are not bad cope with this task, explains Head of Artificial Intelligence in Facebook Yann LeKun (Yann LeCun).

In fact, even looking from the back there are a lot of signs by which we recognize a familiar person. For example, Mark Zuckerberg is easy to learn, because he always wears the same t-shirt.

Neural network trained 40 000 photos from Flickr, some of which face was clearly visible, but on the other - bad. After training system showed recognition accuracy 83%.

Results of the study presented earlier this month at a conference on Computer Vision and Pattern Recognition in Boston.

In the future, the algorithm can be implemented in commercial products Facebook. For example, the program Moments , which performs automatic sorting of archive pictures of individual events, with the recognition of names all the friends in the photos.

So sunglasses and hoods not hide abbr> you from Big Brother,% username%.

Source: geektimes.ru/post/252432/