Facebook has been act upon on facial realization for years to motorcar - rag photographs , but has now hit a point where its technology is ‘ closely go up human - level execution . ’ In fact , in some ways it might even be better .
The social web ’s in - house research squad has been mould on a project calledDeepFace : cutting - edge facial recognition software which maps 3D facial features , turns them into a compressed good example , then filters the result by color to place specific facial elements .
And boy , does it mold . Pitted against a mental test where it ’s ask to identify whether two people in side - by - side image are the same person , it ’s 97.25 percent precise . When human beings take the test , they score 97.5 percent . Good grief that ’s telling . It ’s also taken some time and solitaire to achieve . The scheme uses a staggering 120 million parameters to distinguish cheek , and has been trained using a pool of 4.4 million label faces from 4,030 different hoi polloi on the social web to reach this point .
The feature wo n’t be rolled out on the website just yet , though . In fact , Facebook istaking it to the IEEE Conference on Computer Vision and Pattern Recognitionthis June to see what academic and other research worker make of it . Hopefully more than a few secrecy advocates as well . [ FacebookviaTechnology Review ]
FacebookFacial Recognitionmath
Daily Newsletter
Get the best tech , science , and culture intelligence in your inbox day by day .
News from the future , delivered to your nowadays .
You May Also Like
