A Classification Method for Silhouettes of Various Clothes

Abstract

In this paper, we propose an effective classification method for the silhouettes of various kinds of clothes. There are two approaches to analyzing clothes. The first method focuses on the basic elements of a garment's design, from the viewpoint of its creator. The second method features detailed categories with respect to the garment made. The silhouette of a garment is one of the most important pieces of information in fashion design trends. Here, we focused on classifying silhouettes that lead to the creation of trends, rather than classifying items made of clothing. This classification makes it possible to create a data set of silhouettes that can be used to multi-class classifications in a deep neural network.

Silhouette Classification of Designer’s Collections in Luxury Fashion Brands

“…2. The silhouette classification method proposed in our previous study [12] was adopted. We focused on the 11 positions of the clothing; Total height, Neck, Shoulder, Sleeve width, Waist, Hem high, Hips, Length high, Hem line, Knee, and Length, considering that the feature values are the positions measured by the pattern maker during design.…”

Feature-based fashion flat sketch design using automatic module alignment algorithm

“…Therefore, it was necessary to select representative modules from each category to be stored in the database. For this, Tsuru et al (2019) performed the Ward's method of cluster analysis using the dimensions measured from the clothing image worn by the model. In this study, the bodice modules were grouped with respect to their silhouettes through cluster analysis using the measurement data.…”