Cool work! I'm curious regarding the training dataset. What is the distribution of faces by race/age? Also regarding the raters, what is their distribution? (Race/age/cultural background)?

It's widely known that attractiveness is heavily dependent on cultural upbringing.

Signed, Butthurt dev whose best pic only rated an "OK".

EDIT: You also rated Yoona (a Korean pop star) as just "Nice": http://imgur.com/uJVnQ9S. I guess that makes me feel better about my "OK". I'd stay out of Korea if I were you---I hear their fans aren't very forgiving.

I am Radu, one of the authors.

As mentioned in our article (http://arxiv.org/abs/1510.07867v1), the training dataset for attractiveness is from Blinq. The underage and above 37 years old face images were discarded. All people are heterosexual in our training dataset and mainly from Zurich area, Switzerland. Therefore, our model of attractiveness fits the cultural bias of Zurich, Switzerland.

We consider faces.ethz.ch a little fun tool. I hope that the fans of Yoona will understand :) With more training data from abroad Switzerland our algorithms will fit their expectations.

For age and gender we used much more and diverse data, therefore are more reliable for the majority of ethnic backgrounds.

But seriously, what purpose would it serve over a promise in TOS? A normal user will trust both (except the button would be bullshitting them). A privacy-obsessed one will trust neither.

Edit: Sorry, missed it! http://arxiv.org/pdf/1510.07867v1.pdf

Picture or Extracted Face?