Facebook has released the image identification software it uses to detect faces and objects within users’ photos, the social network has announced.
Facebook has released three main algorithms DeepMask, SharpMask, and MultiPathNet, with the hope that doing so will motivate people to experiment and ultimately help improve its own image recognition technology.
In a nutshell, how it works is that the DeepMask algorithm creates a rough mask of objects in an image. The SharpMask algorithm refines the detail, while MultiPathNet aims to identify what the object masks actually represent.
Currently, when you upload an image to Facebook you will receive suggested tags based on faces that the algorithm thinks it recognises. By releasing its code to the public, Facebook is betting on expanding the software’s capabilities.