Discrimination of clothing materials from smartphone camera images

Information on fabric material is necessary in washing and ironing clothing. However, indication on a care tag may peel off or the tag may come off due to deterioration over time. Discrimination of the material from the fabric itself is not easy for a general person. Estimating the material of an object is one of the challenging tasks in computer vision. This paper deals with the identification of cloth materials using computer vision. We studied a method to discriminate the fabric material from the image of clothing taken by a smartphone camera. First, we investigated the relationship between image resolution and discrimination accuracy using a convolutional neural network (CNN). As a result, we observed that the accuracy changes with resolution and that the resolution at which the accuracy is highest differs depending on the material. Based on these results, we proposed a fabric material discrimination method using multi-resolution images by combining two CNNs. As a result of the evaluation experiment, the proposed method discriminated six kinds of fabric materials with 87.1% accuracy, and the accuracy was significantly higher than that of the comparison method without using multi-resolution images.