We’re introducing Segment Anything Model 2, the first unified model that can identify which pixels belong to a target object in an image or video.