Because that's the kind of image that AI trainers like the most? Would they rather train them on old newspapers?
That would be the "oiled bodybuilder" applied to image training. Maybe similar and clearly defined lighting also allows AI's to match features much better, specially volumes.