Home // AISyS 2024, The First International Conference on AI-based Systems and Services // View article

Towards AI-Generated African Textile Patterns with StyleGAN and Stable Diffusion

Authors:
Christelle Scharff
Samyak Rakesh Meshram
Krishna Mohan Bathula
Fnu Kaleemunnisa
Om Gaikhe

Keywords: African wax patterns; Fréchet Inception Distance (FID); Stable Diffusion; StyleGAN

Abstract:
Wax are traditional colorful textiles worn across Africa. They are composed of patterns of geometrical and symmetrical shapes that repeat indefinitely. This paper explores and compares the generation of African wax designs using StyleGan2-ADA, StyleGAN3 and Stable Diffusion architectures on a curated synthetic dataset of 2000 1024x1024 images obtained with DALL·E 2. The generated wax designs are evaluated using Fréchet Inception Distance (FID). StyleGAN2-ADA and Stable Diffusion generated better images. StyleGAN2-ADA generated designs diverse in colors, shapes and details with some symmetry and repetition. Stable Diffusion was stronger with symmetry and repetition, but it generated less details. By providing a new tool for creating customizable wax designs, this study has the potential to have an impact on the fashion industry. It is novel as it makes a case for inclusive AI by focusing on applications outside the scope of today’s mainstream fashion industry. It also shows that the suggested approaches are promising to produce a variety of plausible and culturally appropriate designs. Our next step is to work with African fashion designers and wax experts to validate the resulting designs.

Pages: 52 to 57

Copyright: Copyright (c) IARIA, 2024

Publication date: September 29, 2024

Published in: conference

ISBN: 978-1-68558-192-3

Location: Venice, Italy

Dates: from September 29, 2024 to October 3, 2024