Yes, of course, what constitutes sensible image recognition is just a matter of opinion, and opinions of some algorithm are as valid as yours. In fact, my pseudo-random number generator seeded with image binary accurately recognizes 100% of images I give it (based on my newly developed definition of image recognition).
I have got no idea of the point you are trying to make. You seem to be criticizing something I said but not sure what. You do realize that I was trying to make a somewhat ironic / somewhat humorous comment that you have to be careful that what you test for is relevant and that one isn't focusing on a narrow weakness which may not actually be that relevant for the general case.
You're making a comment in response to a specific research paper. Therefore, I interpret the comment within the context of that paper. So you're implying that the paper is "focusing on a narrow weakness which may not actually be that relevant for the general case". I disagree.
Any 2d image is an optical illusion, so it makes no sense to criticize human image recognition based on it being 'fooled' by illusions. The real criteria for whether image recognition works well or not is altogether different.
