Facebook's Facial Recognition Algorithm Easily Spots Faces In Photos
Jotham D. Funclara | | Jun 24, 2015 06:08 AM EDT |
(Photo : Reuters) Facebook recently launched PIPER, a facial recognition software that can identify people even when their faces are distorted.
Social media leader Facebook's PIPER facial recognition software can scan photos and identify people even when their faces are obscured or covered, says the Wall Street Journal. This is the next big step for the social media site's AI, and with it, Facebook aims to recognize people with an efficiency as close to the human brain as possible.
Like Us on Facebook
The PIPER facial recognition algorithm uses visual cues like body shape, posture, hairstyle, and even manner of clothing to identify people when their faces are obscured. Facebook and the University of California conducted a study to test the technology by using 40,000 public photos on Flickr. Some photos had the subjects' faces in full view, and others had the same subject with their faces obscured. In the end, the algorithm managed to accurately recognize subjects 83 percent of the time. Facebook claims that this software can easily identify a subject in one picture out of a chunk of 800 million photos in as little as five seconds.
According to Yann LeCun, Facebook's head of AI technology, PIPER uses a vast array of signals to identify one person from the next without fixating on the face. The human brain functions similarly, so that a person can spot the individual differences in the subject's physical traits when compared to the rest of the world, and recognize someone with a full-face cap, a mask, or even in full costume.
LeCun is an expert in computer vision and pattern recognition who was hired by the social media giant in 2013. He previously worked as a Belllabs researcher, where he helped create and develop software technology that allows banks to identify handwriting on checks.
Facebook's latest technological advancement is a double-edged sword for those with privacy concerns. On the one hand, it could easily notify people whenever a photo of them shows up on the social media site, thus preventing the spread of potentially damaging pictures. On the other hand, the fact that Facebook can now identify a person even when they are obviously hiding their identity poses a real threat to their privacy in the first place.
Apparently though, the tech giant isn't finished with facial recognition functions. Facebook also recently launched Moments, a program that randomly scans a user's photo roll, identifies the people in the photos, and offers to share the photos with whomever is in the picture.
©2015 Chinatopix All rights reserved. Do not reproduce without permission
EDITOR'S PICKS
-
Did the Trump administration just announce plans for a trade war with ‘hostile’ China and Russia?
-
US Senate passes Taiwan travel bill slammed by China
-
As Yan Sihong’s family grieves, here are other Chinese students who went missing abroad. Some have never been found
-
Beijing blasts Western critics who ‘smear China’ with the term sharp power
-
China Envoy Seeks to Defuse Tensions With U.S. as a Trade War Brews
-
Singapore's Deputy PM Provides Bitcoin Vote of Confidence Amid China's Blanket Bans
-
China warns investors over risks in overseas virtual currency trading
-
Chinese government most trustworthy: survey
-
Kashima Antlers On Course For Back-To-Back Titles
MOST POPULAR
LATEST NEWS
Zhou Yongkang: China's Former Security Chief Sentenced to Life in Prison
China's former Chief of the Ministry of Public Security, Zhou Yongkang, has been given a life sentence after he was found guilty of abusing his office, bribery and deliberately ... Full Article
TRENDING STORY
-
China Pork Prices Expected to Stabilize As The Supplies Recover
-
Elephone P9000 Smartphone is now on Sale on Amazon India
-
There's a Big Chance Cliffhangers Won't Still Be Resolved When Grey's Anatomy Season 13 Returns
-
Supreme Court Ruled on Samsung vs Apple Dispute for Patent Infringement
-
Microsoft Surface Pro 5 Rumors and Release Date: What is the Latest?