So far, we have seen the image recognition problem, where the model predicted either image at any of the chosen classes or objects, such as is it a car, or perhaps a pedestrian, and so on. In addition to that, we have also seen the object localization and detection, where we put a bounding box to those classes or chosen objects. But for now, that is not important. Anyway, in both the cases, we are interested in establishing whether an image had a certain type of object.