Anyone who has ever had a less than flattering snapshot posted by someone else on Facebook knows the problem. You were at a wedding or a bar mitzvah or a party and were caught in the corner of a casual picture. The picture was posted on Facebook and the person who shared it tagged all of the people in the photo. This means that each person has been formally identified and the picture has been linked back to their page.
But what if you do not like the picture? What if you do not want anyone to know that you were at the place or event? Well you can just untag yourself from the photograph, but sometimes the damage is already done.
If you have tagged photos on Facebook, then you know how it works. Facebook matches people’s faces to their images that are already in the system and makes suggestions for tags to the person who uploaded the picture. But sometimes a person is partially obscured so you cannot tell who it is. Worse even, Facebook may suggest the wrong name and then you end up tagging the wrong name in the picture.
Facebook now has a solution to this problem which, at the same time, creates a whole new set of problems. Facebook’s artificial intelligence lab is now using an experimental algorithm that can recognize people even when their faces cannot be seen in pictures.
Yann LeCun, head of artificial intelligence at Facebook, told NewScientist, “There are a lot of cues we use. People have characteristic aspects, even if you look at them from the back. For example, you can recognize Mark Zuckerberg very easily, because he always wears a gray T-shirt.”
Researchers used 40, 000 photos of people who were both in full view and obscured to develop the new algorithm.
“If, even when you hide your face, you can be successfully linked to your identify, that will certainly concern people, ” said Ralph Gross at Carnegie Mellon University in Pittsburgh, Pennsylvania. “Now is a time when it’s important to discuss these questions.”
But camera shy people who did not wish to be identified now have more to worry about.