The Ensemble-Building Challenge for Fashion Recommendation: Investigation of In-Home Practices and Assessment of Garment Combinations

Fashion is a domain that poses new and interesting challenges for recommender systems. While most recommendation problems seek a single-point solution (e.g. a product the user will purchase), individual garments must function within a wardrobe system, and must ultimately be matched with other garments to build an outfit. The outfit-building challenge is poorly understood in academic literature and professional practice. Here, we present data from two sources: subjective self-reports from consumers about their outfit-building practices, and assessments (by expert and crowd-sourced assessors) of computer-generated outfit combinations pulled from a real-world wardrobe. Results illuminate the objectives and obstacles of consumers in the daily dressing decision, and support the complexity of building combinations from a large set of individual garments.