Face and Eye-ware Classification using Geometric Features for a Data-driven Eye-ware Recommendation System

In this work we present an automatic shape extraction and classification method for face and eye-ware shapes. Our novel eye-ware shape extraction algorithm can extract the polygonal shape of eyeware accurately and reliably even for reflective sun-glasses and thin metal frames. Additionally, we identify key geometric features that can differentiate reliably the shape classes and we integrate them into a supervised learning technique for face and eye-ware shape classification. Finally, we incorporate the shape extraction and classification algorithms into a practical data-driven eye-ware recommendation system that we validate empirically with a user study.