One for the Road: Recommending Male Street Attire

Growth of male fashion industry and escalating popularity of affordable street fashion wear has created a demand for the intervention of effective data analytics and recommender systems for male street wear. This motivated us to undertake extensive image collection of male subjects in casual wear and pose; assiduously annotate and carefully select discriminating features. We build up a classifier which predicts accurately the attractive quotient of an outfit. Further, we build a recommendation system - MalOutRec - which provides pointed recommendation of changing a part of the outfit in case the outfit looks unattractive (e.g. change the existing pair of trousers with a recommended one). We employ an innovative methodology that uses personalized pagerank in designing MalOutRec - experimental results show that it handsomely beats the metapath based baseline algorithm.