Is there a difference in the visual activation in humans and in deep networks when selecting the category of an object?