Within seconds of looking at a picture, the human eye can understand every object in it. Now, Facebook plans to bring that capability to every picture you upload. The firm hopes its image detection*, segmentation* and identification* will one day suggest products for consumers to purchase, identify healthier foods for consumption and provide health stats in real time* — just by analyzing users’ photos. The entire effort is based on “deep learning” and in this project, Facebook puts images through an assembly line using algorithms* that process different components of the image. Sixty artificial intelligence (AI)* scientists have been working to perfect the new algorithms. Systems see pixels as a series of number values corresponding to change in color, but that is where the algorithm comes in. The technology classifies the image and then states what it sees, for example, it could see a person, a dog and sheep and will label each one accordingly. Then it detects the image’s core components, which means the algorithms must learn how to separate important elements of the image from the details in the background or other noise. (SD-Agencies) |