Facebook can recognise you in photos even if it can't see your face
-
Loading... - J. OLIVEIRA
- 24 Jun 2020
- 665 Views
- 0 Like
- 0 Comment
Facebook can already suggest people to tag on photos based on its facial recognition technology, but it’s working on a way to recognise people without relying on it.
An experimental algorithm from Facebook’s artificial intelligence lab can recognise people in photographs even if their faces aren’t visible. The algorithm, reported by New Scientist, looks for other noticeable features like hairstyle, body shape, clothes, and pose to identify a person.
The research team behind the experiment took almost 40,000 public photos from Flickr and ran them through the software. Some of the photos had people with their faces fully visible while others photos had people turned away from the camera.
The final result was the algorithm was able to recognise people with 83% accuracy.
Last week, Facebook released an app in the US which uses facial recognition to identify people. Its Moments app scans your phone’s photos to find friends and moments, and groups them based on who’s in the photo, where they were taken and when they were taken.
The app won’t be released in Europe as there’s no opt-in feature, and Facebook must reach an agreement with the Irish regulator on how to offer the feature to EU residents before it can do so.
In its annual report released today, the Office of the Data Protection Commissioner briefly mentioned the service, saying it had “continuing engagement” with Facebook Ireland before any new feature was introduced last year.
[We had] continuing engagement with Facebook Ireland prior to the introduction of new features, alterations to existing features, privacy settings and controls for individuals, advertising functions and the recent launch of its new “Terms and Conditions” in January 2015.
This included legal and technical examinations of the “user facing elements” of Facebook’s offering, and the organisational and technical processing that goes on behind the scenes. In addition, a substantial on-site review of audit recommendations took place in mid-2014.
