Facebook is known for being creepy due to all its privacy issues, but the social network might seem extra creepy with its new facial recognition technology. There's a strong hate for facial recognition, and we doubt Facebook's implementation will make it any more acceptable.
Facebook's facial recognition software is quite advanced, probably something only the military or the NSA has access to. According to a new report from Facebook, the technology researchers are looking into has the ability to recognize a person's face just as accurate as a human being. If this is real, then the social network is turning into a scary place, and only a drastic change in Facebook's privacy policy and options could allow such a software to move forward.
Bear in mind that Facebook has already implemented facial recognition in its software, you might have noticed it when tagging your friends or family in photos. However, this software is far from accurate, and many times require the user to figure out who person's are, manually.
The social network's new facial recognition software, now known as "DeepFace", is aimed at fixing the accuracy issue, along with recognizing a person even if their face is turned sideways.
So far, DeepFace is able to recognize faces 97 percent of the time, a figure that should give Facebook users the creeps. It also proves how advance this system is, since most facial recognition software requires a clear view of a person's face to know who it is.
How does DeepFace work
The system works its magic by creating a 3-D version of faces within an image, then analyze thee faces using an artificial intelligence system known only as Deep Learning.
Deep Learning is a system that mimics how neurons structure in the brain works, which it uses to analyze large data sets and draw connections to the image in order to pinpoint who the person is.
We've heard a lot about Deep Learning before, and Facebook is not the only company taking advantage of what this system has to offer. Google does it, and so does Netflix in its recent bid to predict what users want watch next.
