Facebook works on many interesting projects such as storing a petabyte of data on blu-rays, and its strides with facial recognition takes the technology even further.
Called DeepFace, the social network says it’s 97.25 percent accurate. To put things into perspective, humans are 97.53 percent accurate when it comes to recognizing faces.
MIT Technology Review explains how it works:
“DeepFace processes images of faces in two steps. First it corrects the angle of a face so that the person in the picture faces forward, using a 3-D model of an “average” forward-looking face. Then the deep learning comes in as a simulated neural network works out a numerical description of the reoriented face. If DeepFace comes up with similar enough descriptions from two different images, it decides they must show the same face.”
“Simulated neural network” is a fancy term for AI software, and it involved some 120 million total connections.
Facebook only used four million photos across nearly 4,000 users in tests, and says it’s improved existing facial recognition accuracy by over 25 percent.
From University of Washington research Neeraj Kumar:
“I’d bet that a lot of the gain here comes from what deep learning generally provides: being able to leverage huge amounts of outside data in a much higher-capacity learning model.”
To learn more about Facebook DeepFace, you can download the research paper here, and the company plans to share its results in June at the IEEE Conference on Computer Vision and Pattern Recognition.