Facebook’s astonishing AI algorithm in Facial Recognition

Pierluigi Paganini June 25, 2015

Facebook developed an astonishing algorithm for facial recognition considered as the latest Artificial Intelligence step, it’s name is PIPER.

Artificial Intelligence has been playing an important role to offer better and outstanding features in applications and a lot of computer giants and agencies have spent huge amount of money to acquire more and more. Facebook has created an AI research laboratory and hired NYU professor “Yann LeCun” as the head of artificial intelligence since 2013. This investment led to a successful experimental algorithm that uses other identifying clues such as hair style, clothing, and body shape in order to recognize people even if their faces are not clear.

This Facebook algorithm is considered as the latest AI step in facial recognition called PIPER. The technology is able to identify people even when faces are obscured. PIPER has been examined by use of a dataset, consisting of over 60,000 instances of 2000 individuals collected from public Flickr photo albums with only about half of the person images containing a frontal face. The final result is quite astonishing and as Facebook said:

“it has the ability to identify individuals with 83% accuracy”.

Facial recognition Facebook

The Pose Invariant Person Recognition (PIPER) method, accumulates the cues of poselet-level person recognizers trained by deep convolutional networks to discount for the pose variations, combined with a face recognizer and a global recognizer.

Facial recognition use has been rapidly increasing both in commercial products, as well as by law enforcement lately. Facebook has used facial recognition technology to enable tag suggestion and for its new photo sharing application named Moments that was introduced on June 15, 2015.Facebook facial recognition software helps users tag photos by suggesting tags of their friends. This software uses an algorithm to calculate a unique number (“template”) based on someone’s facial features, like the distance between the eyes, nose and ears. This template is based on profile pictures and photos users have been tagged in on Facebook. Algorithms like this could help application such as Moments to provide better accuracy and higher user satisfaction.

“If, even when you hide your face, you can be successfully linked to your identiy, that will certainly concern people,” said Ralph Gross at Carnegie Mellon University in Pittsburgh, Pennsylvania. It goes without say that, the algorithm and its ability to identify someone even when they are not looking at the camera is very impressive, but it might bring some serious privacy issues to the fore.

Also, these advanced facial recognition algorithms like PIPER bring limitless possibilities for solving crimes and putting an end to drawbacks of the traditional algorithms which relied on ‘face alignment’ based on locations of the two eyes and have difficulty to function correctly in situation person wearing sunglasses or face is occluded. It’s hoped these researches can be used to help the police act faster and more accurate in crime prevention.

About the Author

Ali Taherian (@ali_taherian) is an enthusiastic information security Officer. He’s finished his education in information security and has recently been involved in banking software and payment security industry. Taherian is proud to be certified IBM Cloud Computing Solution Advisor and ECSA and enjoys sharing and tweeting about security advances and news.

Edited by Pierluigi Paganini

(Security Affairs – facial recognition, Facebook)



you might also like

leave a comment