Multimedia ontology based complementary garment recommendation

TLDR: A novel multimedia ontology based framework to recommend garments to the users for a specific occasion to form a complement outfit pair, which requires dealing with intrinsic subjectivity involved in modeling the domain.

Abstract: In this work, we propose a novel multimedia ontology based framework to recommend garments to the users for a specific occasion to form a complement outfit pair, which requires dealing with intrinsic subjectivity involved in modeling the domain. The framework automatically derives personal traits of a user such as body color, dimensions from her photograph and garment attributes such as color, pattern from input garment to interpret their respective context. Depending on these contexts, system then recommends outfits that are complementary and satisfactory in terms of reference garment color and user personality using semantic web and image processing techniques. The recommendation model compiles knowledge of clothing in an ontology encoded in Multimedia Web Ontology Language (MOWL) which is capable of describing domain concepts in terms of their media properties and utilizes probabilistic reasoning scheme to reason with them. We have experimented and validated our approach with clothing liking of various models on a collection of garments downloaded from various websites, thus providing an effective garment recommendation interface to the user.