Facebook is testing DeepFace, a face recognition program that is 97.5 per cent accurate
Facebook wants to increase revenue by encouraging engagement, so it has invented DeepFace, a feature with a sinister name and the ability to recognise and tag a face in a photo with an accuracy rate of 97.25 per cent. This isn’t taking place in a far-off future: DeepFace is now being trailed in the US as an opt-in feature, although it is not yet useable in the UK.
That means every photo of you put on Facebook, whether flattering or incriminating, could potentially be linked to you by Facebook, even if it never appears on your profile.
Facebook has been developing the feature for over a year. In a paper published last June, Facebook said:
Our method reaches an accuracy of 97.35 per cent on the Labeled Faces in the Wild test dataset, reducing the error of the current state of the art by more than 27 per cent, closely approaching human-level performance.
However, European users of the social network are unlikely to find themselves randomly tagged in the backgrounds of strangers' photos just yet: the European Union currently forbids this kind of tagging.
DeepFace uses technology developed by an Israeli company Face.com, which Facebook bought in 2012.