Article: Composite material surface texture generation and cultural and creative design application based on generative adversarial networks Journal: International Journal of Materials and Product Technology (IJMPT) 2026 Vol.70 No.6 pp.146 - 169 Abstract: This paper proposes a composite material texture generation and cultural and creative design method based on generative adversarial networks (GANs). A multi-scale convolutional generator and self-attention mechanism are constructed to capture microscopic fibre structures and macroscopic texture layouts. Physical constraint encodings of fibre orientation, interlayer interfaces, and micro-defects are embedded in the generator. A structure discriminator evaluates the microscopic realism of the material, while a design style discriminator, combined with a style encoder, applies multi-dimensional latent style vectors to control colour, texture density, geometric feel, and visual contrast. Latent space interpolation and a texture fusion module are used to achieve a smooth combination of different material textures and styles. A design adaptation module performs multi-scale splicing and visual evaluation to output usable texture images for cultural and creative designs. Inderscience Publishers - linking academia, business and industry through research

Title: Composite material surface texture generation and cultural and creative design application based on generative adversarial networks

Authors: Yuandong Jiang; Shuyi Wang

Addresses: School of Design and Art, Xijing University, Xi'an, 710000, Shaanxi, China ' Teaching Experiment Center, Xi'an Academy of Fine Arts, Xi'an, 710000, Shaanxi, China

Abstract: This paper proposes a composite material texture generation and cultural and creative design method based on generative adversarial networks (GANs). A multi-scale convolutional generator and self-attention mechanism are constructed to capture microscopic fibre structures and macroscopic texture layouts. Physical constraint encodings of fibre orientation, interlayer interfaces, and micro-defects are embedded in the generator. A structure discriminator evaluates the microscopic realism of the material, while a design style discriminator, combined with a style encoder, applies multi-dimensional latent style vectors to control colour, texture density, geometric feel, and visual contrast. Latent space interpolation and a texture fusion module are used to achieve a smooth combination of different material textures and styles. A design adaptation module performs multi-scale splicing and visual evaluation to output usable texture images for cultural and creative designs.

Keywords: generative adversarial network; GAN; composite material texture; multi-scale convolution; style control; visual diversity.

DOI: 10.1504/IJMPT.2026.155374

International Journal of Materials and Product Technology, 2026 Vol.70 No.6, pp.146 - 169

Received: 12 Jan 2026
Accepted: 10 Apr 2026
Published online: 30 Jul 2026 *

Title: Composite material surface texture generation and cultural and creative design application based on generative adversarial networks

Keep up-to-date