Facial recognition technology is getting really good. Like, creepy good. This makes some people uncomfortable. Adam Harvey is one of those people, so he has started a project to try and fool automated eyes with creative uses of things like makeup and hairstyles.
The CV Dazzle project is named after an old military program designed to keep cover intact for war ships. The basic concept is to disrupt the basic identifiers that tell a computer that it’s looking at a face. It’s interesting to see what works and what doesn’t. Some simple additions make the programs confused, while other complex attempts are easily seen right through.
The ultimate goal is software that can analyze your face and generate what he calls an “anti-face” that can be applied to flummox the cameras. It’s an interesting concept and it doesn’t seem crazy that people would eventually adopt something like this when they don’t feel like being tracked. It seems like it could also be applied in simpler ways, like a pair of facial recognition-killing sunglasses.
The visualization of the OpenCV facial recognition software is below and it’s fascinating to see how it work, relentlessly analyzing pieces of the frame to see if there are indicators that it’s looking at a face.
It even works with things like Facebook, which is getting better all the time at recognizing our mugs. What do you think? Would you consider doing something like this to keep your face out of a data base? Is this something people might use when they’re up to no good?