Fashion-Sketcher: A Model for Producing Fashion Sketches of Multiple Categories

Abstract

Fashion sketches play a critical role in the initial stages of fashion product design. This situation has motivated the development of artificial intelligence (AI) techniques for automatically generating fashion sketches. We present Fashion-Sketcher, a hybrid deep generative model able to generate multi-class fashion sketches through two stages. At the first stage, we design a Contour Generation Network to synthesize contour images with a given categorical vector. At the second stage, we design a Sketch Translation Network to translate the contour images to the sketch images by extending the StyleGAN2 model to the conditional version. The quantitative and qualitative analysis demonstrates that our method is capable of synthesizing high-quality fashion sketches.