Fashion coordinates recommender system using photographs from fashion magazines

摘要: Fashion magazines contain a number of photographs fashion models, and their clothing coordinates serve as useful references. In this paper, we propose recommender system for using full-body from magazines. The task is that, given photograph item (e.g. tops) query, to recommend other items bottoms) that appropriate the query. With proposed method, use probabilistic topic model learning information about visual features in each region. We demonstrate effectiveness method real magazine two style sharing services with making top (bottom) recommendations bottom (top) photographs.