Humans see faces everywhere. We see a face in the craters of the moon, in wall sockets, sideways in punctuation :-) and just about anywhere else two dots and a line are arranged in even approximately the same positions as two eyes and a mouth.
Once we recognize something as a face, we process it differently from other visual stimuli. Certain parts of the brains are triggered preferentially by faces. We are especially good at perceiving faces: we can pick out matching faces faster than matching abstract patterns, and distinguish non-matching faces more easily than other images. This only works, however, when our brains recognize the faces as faces: if you flip faces upside-down, they no longer trigger the “face” switch for us, and we become much worse at distinguishing them. The same thing happens if you digitally scramble facial features, so that there’s an ear in the middle and an eye on the chin and a mouth slanting across the forehead, or any other mix-up that makes the face no longer be arranged like a face. Our brains are specialized to perceive face-shaped patterns much better than other patterns.