Classification of the category of clothing item after bringing it into limited shapes

In this paper, we present a method to classify the category of clothing item. When visually recognizing clothing items based on their shapes, two types of shape variations make the recognition difficult: physical deformation and the variety of size and design. In order to lessen the effect of deformation, first, the item of interest is reshaped into one of a small number of limited shapes. To absorb the variety of size and design, we propose to represent each limited shape using a likelihood image, which is made by combining clothing contours of various design items in the limited shape. Experimental results using long-sleeve shirts and trousers showed a good prospect of the proposed method.