Would you believe Facebook's facial-recognition software is better than the FBI's? Well, it is, and it makes a lot of sense.
The FBI's facial-identification software, Next Generation Identification - which is rolling out later this year - has its image databases stocked entirely by law enforcement. That gives it a grand total of about 50 million images.
According to early reports, NGI takes 50 guesses on every face and is only 85 percent confident it's right.
Facebook's DeepFace system, on the other hand, has a pool of 250 billion - yes, billion - photos. Given that Facebook has more than 1 billion users, that makes sense. Still, that's a lot of photos.
It also means that Facebook can match up faces with 97 percent accuracy.
As it stands, facial recognition is still a brand-new surveillance tool. Most modern security cameras are angled downward, making it difficult for facial-recognition programs -- both Facebook's and the FBI's -- to get a clear look at you.
What makes DeepFace unique is that if the software can't get a hard match from looking at your face, it moves on to your friends. Most people have a Facebook, and almost every Facebook page has a picture of someone's face on it.
By hunting down your friends' Facebook pages, DeepFace is able to identify you without requiring a 100 percent match on your photo.