DEFINING THE MAIN FEATURES OF CLOTHING TO APPLY DEEP LEARNING IN APPAREL DESIGN

摘要

The paper is devoted to defining the features of clothing to apply deep learning in apparel design. The images of women's outerwear were selected with the help of reverse image search. The images of women's duffle coats, coats and suit jackets were selected. The selected material was sampled for the next categorical principal components analysis and general assessment of differences that are caused by specific features of garments. It was revealed that similarity search might be used to perform the selection of models to define the typical design solutions. However, a process of defining the solutions cannot be automated yet. The indicators of clusters, which were revealed in the result of the categorical principal components analysis, define the structure of the database for the deep learning. Based on the results of performed online survey, it was considered advisable to use as labels specific features of the particular garment type rather than its name. Each label refers to one of the main features of a garment type.