The face recognition system of Facebook is becoming more and more curious. After recognizing facial features to identify a person, the social network is now interested in everything that a photo can contain, such as the environment or any recognizable object in the image.
"Until recently, online research was limited to text, even when looking for images," says Joaquin Candela, head of Facebook's artificial intelligence. "For an image to go up, it had to be properly tagged or have the right legend, until now." Because now, Facebook can take care of itself by knowing what is in a photo: the people present, as it was already since the establishment of facial recognition, but also the place where the photo has Taken according to what is seen there, or the different objects that can be found there.
"In return, this allows us to better describe the photos for the visually impaired," says Joaquin Candela, "in addition to providing better search results for photos and videos." This will result in the addition of 12 actions that the system can use to describe the photos. Thus, Facebook will be able to find photos of people dancing, walking or riding if it recognizes these actions in a photo.
0 comments:
Post a Comment