Not quite the same, show humans one picture of a giraffe and most humans will recognise the animal in different conditions, poses, angles and so on. Any neural net (up till now) has a difficult time doing the same.
Incorrect. Recognition applications have been able to track objects across disparate frames of video even when obscures from camera view, for at least a decade now [1].