Facebook AI can spot a face 97.25 percent of the time

Researchers from Facebook (tick: DeFacebook, Book of Faces) said that advances in the relatively new artificial-intelligence field known as deep learning have meant that they can spot a face 97.53 percent of the time.

If you show a person two unfamiliar photos of faces show the same person, a human being will get it right 97.53 percent of the time. New software developed by researchers at Facebook can score 97.25 percent on the same challenge.

This is a big step up from previous face-matching software, and is being touted as a success story for deep learning. his area of AI involves software that uses networks of simulated neurons to learn to recognise patterns in large amounts of data.

Yaniv Taigman, a member of Facebook’s AI team, said the error rate has been reduced by more than a quarter relative to earlier software that can take on the same task.

Dubbed DeepFace Facebook’s new software runs a routine called facial verification; it recognises that two images show the same face. It does not do facial recognition yet but some of the underlying techniques could be applied to that problem.

DeepFace remains purely a research project for now. Facebook released a research paper on the project last week, to get feedback from the research community.

DeepFace corrects the angle of a face so that the person in the picture faces forward, using a 3D model of an “average” forward-looking face.

The deep learning kicks in and a simulated neural network works out a numerical description of the reoriented face. If DeepFace comes up with similar enough descriptions from two different images, it decides they must show the same face.