Read the full-text of this article for free

Title: Algorithm for visual communication design: generating textures of composite materials using artificial intelligence

Authors: Dong Chen; Qing Li

Addresses: Department of Comprehensive Arts, Sahmyook University Korea, Seoul, 01795, Seoul, Korea ' School of Fine Arts and Design, University of Jinan, Jinan, 250000, Shandong, China

Abstract: In response to the lack of authenticity and quality in composite material texture generation in the field of visual communication design, this article adopts a self-attention generative adversarial network (GAN) algorithm to study it. This article enhances the texture dataset for composite materials, thereby expanding the available datasets. The self-attention mechanism is applied to the generator and discriminator of generative adversarial networks, and the networks are jointly optimised. Then, the perceptual loss function and an adaptive adversarial loss function are applied to improve the similarity and authenticity between the generated and real images and a progressive fine-tuning strategy is adopted to train the model. The evaluation results of the self-attention generative adversarial network show excellent performance in three indicators, SSIM, PSNR, and perceptual loss, with mean values of 0.82, 23.22, and 0.012, respectively, and the score of the model is 67. Overall, the improved network described in this article offers significant advantages for composite material texture generation, providing a more reliable solution for the field of visual communication design.

Keywords: generative adversarial network; GAN; self-attention mechanism; composite texture generation; visual communication design; image quality evaluation.

DOI: 10.1504/IJMPT.2026.155366

International Journal of Materials and Product Technology, 2026 Vol.70 No.6, pp.27 - 44

Received: 13 Nov 2025
Accepted: 01 Apr 2026
Published online: 30 Jul 2026 *