New Facebook Software Uses AI to Match Faces with (Almost) Human Accuracy
March 18, 2014 4:35 PM
comment(s) - last by
The software is called "DeepFace"
Facebook has created a piece of software that can tell whether two faces match with near-human accuracy. According to
MIT Technology Review
, the new Facebook software is called DeepFace, and it utilizes deep learning for facial verification.
Deep learning is a new branch of artificial intelligence (AI) that uses networks of simulated neurons to recognize patterns in large quantities of data. And facial verification is different from
, because it recognizes that two images reveal the same face where the latter puts a name to a face.
What's more is that DeepFace can accomplish facial verification whether the person is facing forward or not. This is how it works: DeepFace uses a 3-D model of an “average” forward-looking face to adjust the angle of the face it's attempting to verify. This makes the face forward-looking. Then the simulated neural network configures a numerical description of the reoriented face, and if DeepFace finds descriptions from two different images that nearly match, it concludes they must be the same face.
DeepFace [SOURCE: MIT Technology Review]
In testing, the researchers put DeepFace up against a standard data set used to benchmark face-processing software.
When asked if two photos of faces are that of the same person, a human will answer correctly 97.53 percent of the time. DeepFace managed to get it right 97.25 percent of the time with the same challenges.
The report added that changes in lighting will not affect DeepFace's accuracy.
Facebook was able to make such a leap in AI because of its enormous database of photos. It was able to train DeepFace to recognize patterns with its copious amount of photos to choose from, where people are presented in many different poses in various environments within the photos. This made DeepFace more accurate over time.
DeepFace was made by Yaniv Taigman and other members of Facebook’s AI team. They will present their work at the
IEEE Conference on Computer Vision and Pattern Recognition
in June 2014.
MIT Technology Review
This article is over a month old, voting and posting comments is disabled
3/18/2014 5:02:06 PM
"I'd be pissed too, but you didn't have to go all Minority Report on his ass!" -- Jon Stewart on police raiding Gizmodo editor Jason Chen's home
Facebook Prompts More Privacy Anxieties with Facial Recognition Feature
June 8, 2011, 10:28 AM
Quick Note: Whoops, Microsoft Pushed Unwanted Windows 10 to Some Users
October 15, 2015, 9:04 PM
Quick Note: Windows 10 Insider Preview Build 10565 Fixes Boot Camp 6.0 Issues
October 13, 2015, 11:39 AM
Microsoft's HD-500 ("Display Dock"), the Magic Sauce Behind Continuum
October 6, 2015, 5:30 PM
Quick Note: Windows 10 Hits 110 Million Devices, VMs
October 6, 2015, 4:30 PM
Windows 10 on Raspberry Pi, IoT Devices Sees Developer Debut
August 12, 2015, 2:41 PM
Sony Issues Bizzare "Do Not Update" Edict to VAIO PC Owners
August 11, 2015, 9:42 PM
Latest Blog Posts
Sceptre Airs 27", 120 Hz. 1080p Monitor/HDTV w/ 5 ms Response Time for $220
Dec 3, 2014, 10:32 PM
Costco Gives Employees Thanksgiving Off; Wal-Mart Leads "Black Thursday" Charge
Oct 29, 2014, 9:57 PM
"Bear Selfies" Fad Could Turn Deadly, Warn Nevada Wildlife Officials
Oct 28, 2014, 12:00 PM
The Surface Mini That Was Never Released Gets "Hands On" Treatment
Sep 26, 2014, 8:22 AM
ISIS Imposes Ban on Teaching Evolution in Iraq
Sep 17, 2014, 5:22 PM
More Blog Posts
Copyright 2016 DailyTech LLC. -
Terms, Conditions & Privacy Information