The object-detection algorithm, called RetinaNet, scans the plate, identifies the types of food on it and places a frame around each item.