Facebook's facial verification software reaches 'near human-level performance'
The program can match photos of the same individual with 97.25 per cent accuracy - just a shade below human performance of 97.5 per cent
Tuesday 18 March 2014
Facebook has announced that the latest iteration of its facial verification software has reached near “human-level performance”.
The social network’s deep learning algorithms (computer programs that can learn to recognise patterns from large amounts of data) can now say when two images show the same individual with 97.25 per cent accuracy. By comparison, humans get this task right 97.5 per cent of the time.
The new software is called DeepFace, and analyses photos regardless of lighting, expression or whether or not the individual is facing the camera. It does this by creating a 3D map of the facial features in the picture and "recognises" the individual by calculating various numerical descriptions of the face – eg, the ratio between eyebrows and mouth and the distance between nose and ears.
The screen shot above shows the technology in action. Beginning with a snapshot of an individual face (a), the software processes this into a wireframe model (e) and finally a 3D recreation of the face (h).
DeepFace is technically a facial verification system (matching images to images) rather than a facial recognition program (matching images to a name), but many of the underlying techniques are transferrable between the two goals.
Facebook says that it used a pool of 4.4 million labelled faces from 4,030 different people on its network to load the system. The software is not currently being introduced to Facebook itself, but is simply being presented to garner feedback from other researchers.
The social network first introduced its facial recognition software back in 2010 to American users before bringing it worldwide in 2011. In 2012 the EU forced Facebook to drop the functionality in Europe and delete all the facial templates it had collected from users. Facial recognition remains unavailable for users in the UK.
Life & Style blogs
Jennifer Lawrence nude pictures leaked: Reddit removes 'The Fappening' board dedicated to sharing naked pictures of celebrities
iPhone 'Wave': iOS 8 hoax claims you can charge your iPhone in the microwave - you can't
London Fashion Week Spring/Summer 2015: Distraction could not disguise a distinct lack of focus
A bottle of wine a day is not bad for you and abstaining is worse than drinking, scientist claims
iOS 8 HealthKit catches a major bug, forcing Apple's fitness apps to lie down and take it easy
Scottish independence referendum: A nation divided against itself
Scottish referendum results: Cross-party consensus collapses amid Tory-Labour spat on the 'English question'
Scottish independence: David Cameron is becoming the 'George Bush of Britain'
Hilary Mantel 'should be investigated by police' over Margaret Thatcher assassination story, says Lord Bell
Plebgate MP Andrew Mitchell called officer a 'little s**t', claim court documents 'exposing ex-Chief Whip's 'record of abusing police'
Archbishop of Canterbury admits doubts about existence of God
- 1 Rihanna 'nude photos' claims emerge on 4Chan as hacking scandal continues
- 2 Frank Lampard equalises for Manchester City against Chelsea: how the internet reacted
- 3 Stamford Hill council removes 'unacceptable' posters telling women which side of the road to walk down
- 4 Kim Kardashian 'nude pictures' leaked on 4chan weeks after Jennifer Lawrence 'The Fappening' scandal
- 5 Hitler’s former food taster reveals the horrors of the Wolf’s Lair
iJobs Gadgets & Tech
£40000 - £50000 per annum: Ashdown Group: Service Delivery Manager Kingston Up...
£24000 - £30000 per annum + benefits: Ashdown Group: Graduate BI Consultant (B...
£40000 - £55000 per annum: Ashdown Group: Service Delivery Manager (Product Ma...
£50000 - £70000 per annum: Ashdown Group: Cloud Product and Solutions Marketin...