Now, we are going one step further by uniquely identifying the object itself, so not just the type, but we are interested in the unique object:

So we will not just identify whether the image is a car or not, but rather, will also find out if it is specifically my car, your car, or someone else's car. And for animal detection, we may identify someone's cat or dog, not just if the image has a general dog entry. This is the object verification problem; face verification is not different, merely the logic is extended to the human face, and it may be a bit more sensitive, since identifying the wrong person can have a lot of consequences, but the logic stays the same.

The question here is not simply whether the object in the image is a human or pedestrian, but rather whether it is a person with ID X, or perhaps a company employee with some specific number: