Lightweight GAN and Stable Diffusion in batik design generation: a review

Darmin, Darni and Mohd Yassin, Ihsan and Zabidi, Azlee and Osman, Fairul Nazmie and Md Tahir, Nooritawati (2026) Lightweight GAN and Stable Diffusion in batik design generation: a review. Journal of Electrical and Electronic Systems Research (JEESR), 29 (1): 20. pp. 160-167. ISSN e-ISSN : 3030-640X
Identification Number (DOI): 10.24191/jeesr.v29i1.020
Abstract

Generative artificial intelligence has emerged as a transformative tool for traditional textile innovation, including batik design synthesis. Among dominant paradigm, Lightweight Generative Adversarial Networks (LGANs) and diffusion-based models such as Stable Diffusion represent two contrasting architectural approaches in batik image generation. However, their relative suitability for batik motif generation remains insufficiently systematized. This paper presents a comparative review of LGAN and Stable Diffusion in the context of batik design. The analysis examines architectural mechanisms, training stability, computational efficiency, optimization strategies, and evaluation protocols, incorporating quantitative metrics including FID, IS, PSNR, SSIM, LPIPS, KID and CLIP-score. The review indicates that LGANs reduce computational requirements once trained, enabling real-time experimentation with batik motif variations on less powerful hardware, whereas Stable Diffusion demonstrates greater semantic controllability and high-resolution motif synthesis through latent denoising processes. The study further argues that quantitative metrics alone are insufficient to assess culturally grounded textile generation, emphasizing the need for integrated human-centred and authenticity-aware evaluation. This review outlines the basis for a domain-specific evaluative framework to guide architecture selection and future research in sustainable AI-assisted batik innovation.

Item Details
Edit Item
Edit Item
Downloads & Files
[thumbnail of 149170.pdf]
Text
149170.pdf
Download (567kB)
Indexing & Metrics