How to hide from face-detection technology
http://whatsnext.blogs.cnn.com/2012/04/29/how-to-hide-from-face-detection-technology/?hpt=hp_c3
It pains me somewhat to have not thought of this before reading the article, but there you go. Sometimes the obvious only becomes so once articulated and this is one of those cases, at least for me.
Just like the whole CAPTCHA thing, there will be an on-going arms race between the designers seeking to defeat the facial recognition systems and the programmers seeking to account for variation and I doubt there will ever be a lasting winner on either side. I do think that the ‘war’ is in its early stages as I suspect the programmers haven’t been giving a lot of thought to people actively attempting to defeat the algorithms, but the cat is out of the bag now and I am sure that even common criminals (apparently facial recognition in England (London in particular) is a very standard thing) can apply makeup well enough to defeat the current algorithms.
It seems that nearly all these algorithms work on the assumption that the image is a plane (i.e., is two dimensional) and as such it is fairly easy to trick them. I have pondered how I would try to build a 3d representation of the view from the 2d images, but sadly (but not surprisingly) no one seems to care about my ideas (I have tried to read several computer vision books and get lost in the math generally in the first chapter, so those people who are good at groking the current paradigm would naturally consider me a moron). It would be interesting to do that sort of research, but it won’t happen on my own time, I already have too damn many projects to take on something like this.
