Facebook, the social web giant, is working on a new image recognition tool that can recognise the identities of people in photographs even when their face is partially or completely hidden.
Created by Facebook’s artificial intelligence lab, the tool instead of searching and recognising just the face of the user, tracks their unique characteristics such as hairdo, clothing, body shape, and pose as well to confirm identity, reports New Scientist website.
“There are a lot of cues we use. People have characteristic aspects, even if you look at them from the back,” explained Yann LeCun, Head of Artificial Intelligence at Facebook. “For example, you can recognise Mark Zuckerberg very easily, because he always wears a gray T-shirt.”
The team pulled roughly 40,000 public images from Flickr including some images showing full faces and some showing partially covered faces to provide data to build its recognition algorithm. Later tests showed the algorithm to be 83 percent accurate. Notably, the AI was showcased earlier this month during the Computer Vision and Pattern Recognition conference in Boston, Massachusetts.
LeCun said such technology could be used in Facebook’s standalone Moments photo app that was announced a week ago. The app apparently uses facial recognition technology to group photos based on the friends who are in them, similar to the “tagging” used by the Facebook social network. LeCun added it could even be used to help people identify when they’ve been tagged or listed somewhere online.
The new Facebook algorithm however also has a downside when dealing with the privacy. Using the technology, anybody can log in to other user’s account with a hidden face. “If, even when you hide your face, you can be successfully linked to your identify, that will certainly concern people,” said Ralph Gross, a postdoctoral researcher at Carnegie Mellon University. Facial recognition technology itself has been coming under fire recently from privacy advocates.