“In modern face recognition, the conventional pipeline consists of four stages: detect => align => represent => classify. We revisit both the alignment step and the representation step by employing explicit 3D face modeling in order to apply a piecewise affine transformation, and derive a face representation from a nine-layer deep neural network.”
Take a moment to read that again. This is the opening sentence to Facebook’s statement at the Conference on Computer Vision and Pattern Recognition regarding their new and insanely complicated software: DeepFace. To put that robot language into layman’s terms, DeepFace is basically a series of algorithms that can accurately identify a face regardless of the lighting, blurriness, or camera angles. So even though you may think untagging that picture of you chugging a bottle of Skol means it’s not associated with you anymore, Facebook knows exactly who you are and will store that information in its massive database forever.
According to Forbes, humans have a 97.53 percent accuracy rate when it comes to recognizing a face. DeepFace has perfected its software to a 97.25 percent accuracy rate, which means that impressive split second of facial recognition we take for granted has been condensed into a series of algorithms. Though Facebook is already pretty adept at suggesting tags for photos, this new software is much more precise.
So how does it work?
The first step is to change the angle of the photo so that the person in the picture faces the camera, and then a neural network matches the image to similar photos of your face in order to narrow down the search. This neural network is called “Deep Learning” which has nine layers of simulated neurons with over 120 million connections between them. So far, Facebook has taken four million photos of faces belonging to 4,000 people and is using this small scale to train larger capacity software.
Sounds like a lot of hooplah, but it only gets weirder from here. I’m sure you all have noticed how quickly Facebook responds to your other Internet searches. Spend five minutes browsing a pair of sneakers of Foot Locker’s website or order a pizza online and the next time you log on you can bet there’ll be a Domino’s or Nike ad somewhere in your sidebar. Mark Zuckerberg wants to take this type of information gathering to a new level, with DeepFace and another type of software that can analyze your status updates and comments to interpret your mood and context.
These break through computer programs will be able to understand almost everything about you in seconds, so that drunken comment you left on your ex-boyfriend’s photo will give Facebook immediate insight on your recent breakup.
The second scary part about this lack of privacy is its probable existence outside of Facebook. Imagine perusing the aisles at Target while a security camera tracks your every move and stores your specific purchases into a database. If Facebook allows such a software, it’s only a matter of time before other institutions employ it as well.
The thought of all this makes me want to dump my computer into a garbage can, Ron Swanson-style, but I can’t deny the creepy awesomeness of this technological feat. Ten years ago we were using cellphones as hefty as a brick, and computers moved at a snail’s pace. Today we’re constructing software almost as intelligent as the human mind.
When we imagined this stage of the future as little kids, our sticky fingers drew cars that could fly and robots that would respond to our every beck and call. We’ve finally reached this mysterious future decade that we often speculated about, and it’s a vastly different from what we imagined. Fortunately, these futuristic programs are still developing, and we can only hope Facebook notifies the public of these changes in relatable human terms so we know exactly what we’re getting ourselves into.
Featured photo courtesy of: NicNacPhotography
