Given an image, returns the positions of facial features (eyes, nose, lips etc) for each face in the image.