A method of classifying crumpled clothing based on image features derived from clothing fabrics and wrinkles

This paper describes a method of clothing classification using a single image. The method is intended to be used for building autonomous systems, that can recognize casually thrown ordinary clothing. A set of Gabor filters is applied to an input image, and image features invariant to translation, rotation and scale are then generated. In this paper, we propose descriptions of the features, focusing on clothing fabrics, wrinkles, and cloth overlaps. In addition, to deal with situations involving clumped clothing, the description is extended by combining with superpixel representation. Experiments using a state description and classification using real clothing demonstrate the effectiveness of the proposed method.