ColorwAI: Generative Colorways of Textiles Through GAN and Diffusion Disentanglement

Abstract

Colorway creation is the task of generating textile samples in alternate color variations maintaining an underlying pattern. Selecting a colorway is a complex creative task, responding to client and market needs, technical and cultural specifications, and personal artist style. We introduce a framework, “ColorwAI", to tackle the generative task using color disentanglement on StyleGAN and Diffusion while maintaining minimal shape alteration. We present a variation of the InterfaceGAN method for semi-supervised disentanglement, ShapleyVec, which uses Shapley values to subselect salient dimensions from the detected latent direction. Moreover, we present a framework to employ common disentanglement methods on any architecture with a semantic latent space, and test it on DDM and StyleGAN2-ADA. Our results show that StyleGAN’s W space is the most aligned with human notions of color in terms of vector similarities and generated colorways. Finally, we suggest that disentanglement can solicit a creative system for colorway creation, and evaluate it through expert questionnaires and within the lens of creativity theory.