When I say the words “facial recognition,” your brain probably instantly jumps to scenes from crime dramas where you scan a photo into the computer, and the computer proceeds to search through a billion photographs of every human ever born until it comes up with a solid match. But it’s rarely this simple.
Facial Recognition Technology Advances
Believe it or not, most things humans can do with ease, AI struggles with immensely. Have you ever seen robots attempt to walk, or even crawl, in a room littered with obstacles? It’s laughable how they can’t match the capabilities of even the youngest mobile child. Thus, it should come as no surprise that, unlike humans, computers are awful at facial recognition. The technology has slowly been advancing over the years, the latest being Apple’s current facial recognition technology, Face ID, in their most recent iPhone. But in reality, computers have a lot of difficulty with the types of classification that humans do without even thinking.
In approaching the facial recognition technology problem, researchers first had to teach computers how to identify faces. This may seem straightforward enough but think of all the different iterations faces have. You could be looking at a profile a partially obscured face or a face with distortions such as makeup or sunglasses. While humans instinctively still recognize all of these faces as faces (and can recognize people’s faces as being the same person across different iterations), getting machines to recognize faces in all of their forms was a feat in and of itself. Let alone teaching computers to assign an identity to faces.
Facial recognition technology has come a long way since then and has seen many different methods and iterations. The way Apple’s Face ID tech works is the culmination of years of research, painstaking programming, and significant advances in technology that we will attempt to simplify and explain here. Basically, it utilizes an infrared camera to map the specific geography of your face.
Now the first part of that sentence we need to unpack is infrared. It’s probably a term you’ve heard before but aren’t entirely sure what it is. Infrared is a frequency of light that isn’t visible to humans but corresponds to energy emitted by atoms. I can tell I’m losing you already, so here it is in simple terms – infrared lets you visualize heat. The warmer something is, the brighter the infrared rays. This is how night-vision technology works – translating infrared (heat) waves into visual images (which is why in night-vision videos, you’ll see vaguely human-shaped blocks of light moving across the image. Because humans emit a lot of heat).
So back to Face ID, the infrared camera takes a reading of your face’s specific infrared (heat) output. This way, it won’t be tricked by any of the “disguises” mentioned above. It’s reading your face’s specific heat image rather than any superficial visual cues. You could grow a beard, put on some wacky new glasses, slather your face in makeup, etc., and the iPhone would still recognize you. In fact, in some ways, this technology is superior to human facial recognition because you could wear so much makeup that you’re unrecognizable to your friends, or you could put on a mask, or anything else, and your iPhone would still recognize you. Pretty neat for facial recognition technology.