On the potential for facial attractiveness as a soft biometric

This paper describes the first study on whether human facial attractiveness can be used as a soft biometric feature. By using comparative soft biometrics, with ranking and classification, we show that attractiveness does have the capability to be used within a recognition framework using crowdsourcing, by using groups from the LFW dataset. In this initial study, the Elo rating system is employed to rank subjects’ facial attractiveness based on the comparative descriptions. We will show how facial attractiveness attributes can be exploited for identification purposes and can be described in the same way and can add to performance of comparative soft biometrics attributes. Attractiveness does not appear to be as powerful as gender for recognition. It does however increase recognition capability and it is interesting that a perceptual characteristic can improve performance in this way