ReRe♥MaMa

Personalized Aesthetic 

Photo Selection System

  In this work, the personalized aesthetic photo selection system is proposed. There are two key criteria are explicitly considered for the proposed system. One criterion is to select photo by the facial characteristics of the user preference (e.g., face expressions and face shape). The other criterion is to select photo based on aesthetic clothing adapting to personal facial characteristics (e.g., skin tone and face shape).

    To learn a face-adaptive model and comprehensively evaluate it, we collect a large clothing dataset from Lookbook and Chictopia.com, the social networking websites for fashion bloggers. On the websites, bloggers upload all kinds of outfit photos designed to draw attention to their fashion choices. The peers can give their assessment as a form of social interaction. The votes from the peers are taken to wight the score of the clothing matching.