A Look Into Facebook's Potential to Recognize Anybody's Face
npr.org
npr.org
In particular, a conversation about recognition algorithms for sorting pornography at the age-level was brought up. Now, I know this has been covered on here before, but what a massive amount of incriminating information must be kept in those servers.
I really would like to be able to have access to facial model information for my hypothetical advertising company. Can you imagine if I could, for instance, find out what primary five face shapes people in Iowa find trustworthy, with a high (enough) level of accuracy? I'd pay a whole lot for that, hypothetically.
Seems thats not the case.
That being said, if the database is small enough (e.g., in a limited scenario or by applying other non-vision filters first), then the state of the art methods do use a similar approach. However, in practice people use methods with much stronger priors than KNN. Because faces fall into a low-dimensional manifold[1], you can take advantage of that to constrain queries much more than with generic KNN.
[1] My PhD advisor has published several works about this topic, e.g., http://vision.ucsd.edu/kriegman-grp/papers/ijcv98.pdf http://virtualhost.cs.columbia.edu/~belhumeur/journal/invari... http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.123...
...oh wait. Fuck. :(
It's only 15mins long but you can skip to 2:35 for one of the face experiments and to 8:00 for an experiment they're running now. That latter experiment is quite creepy but I'd love to know the results.