Generative Adversarial Networks and Textile Design: Towards an Augmented Craftsmanship in the Context of Digital and Sustainable Fashion

Abstract

This contribution investigates the application of Generative Adversarial Networks (GANs) to the generation of textile patterns, aiming to evaluate the technical and design feasibility of an AI-driven approach in the context of digital and sustainable fashion. A hybrid visual dataset was constructed, comprising historical ornamental motifs and contemporary patterns, carefully selected to ensure morphological coherence and modular repeatability. This dataset was used to train a GAN model on Google Colab, monitoring the visual evolution of the generative outputs at different stages of training. Preliminary results show the emergence of recognizable
structures potentially applicable to textile design; however, technical issues persist, such as digital noise, edge discontinuities, and insufficient resolution for fabric printing. These findings indicate the need for architectural and parametric optimizations, as well as specific evaluation criteria for seamless patterns. The study also highlights the importance of considering the computational costs of generative models from a sustainability perspective, outlining future directions aimed at improving visual quality, scalability, and integration into the industrial fashion design workflow.