Digitale Phrenologie oder so ähnlich, aber wird wohl geglaubt und gekauft. Wohl wegen "KI".
@jakob This thing seems to only evaluate the visual presentation. I'm pretty sure the training data was not without bias, maybe from manual classification.
It's also possible that the training set contains backgrounds that now slips into the evaluation.
Bottom line: No one really knows, what this software finds important.
Hey, how about calling ML systems "automated stereotype reproducer"? "ASR". No "AI" :)