iDesigner: making intelligent fashion designs

Xiaoling Gu, Qiming Yao, Xiaojun Gong, Zhenzhong Kuang

Abstract: This paper presents iDesigner, a novel AI-assisted design system tailored to support intelligent fashion designs. Our proposed system aims to assist fashion designers by automatically synthesizing high-quality product images conditioned on category attributes and texture examples. Since fashion sketches are the fundamental basis of fashion designs, we implement iDesigner with two design assistants, namely Fashion-Sketcher and Style-Transfer. Specifically, Fashion-Sketcher generates a variety of realistic fashion sketches conditioned on the category attribute by mimicking human painters that first draw the outlines and then finish the detailed contents of the objects. Style-Transfer synthesizes the fashion product images by applying texture examples onto the synthetic sketch images with a feature transformation scheme. We validate our approach using a new dataset and demonstrate that our proposed iDesigner can not only successfully generate diverse sketch images conditioned on category attributes, but also generate high-quality fashion product images conditioned on sketch images and texture examples.