Innovative design of ceramic products based on extension reasoning and generative adversarial networks
Year: 2025
Type: article
Abstract: This paper proposes a novel framework integrating extension reasoning with generative adversarial networks to address rigid cultural element expression in ceramic product design. The methodology comprises three innovations: 1) a primitive model analysing object-element, relation-element, and feature-element for topological pattern transformation; 2) multimodal fusion of geometric features, chromatic attributes quantified by GIE-Lab values, and semantic associations; 3) a lightweight network optimised with leaky rectified linear unit activation and Adam optimiser, achieving 30% faster convergence than stochastic gradient descent. Experiments on 3,262 Jingdezhen ceramic images demonstrate superior performance with inception score 5.7 and Fréchet inception distance 2.13, outperforming autoencoder (2.3/5.61), variational autoencoder (3.8/4.54), and deep convolutional generative adversarial network (4.6/5.22). The framework enhances design efficiency by eight times compared to manual processes, validated through Kano model analysis with 89% user satisfaction. This work establishes a technical paradigm for intelligent modernisation of traditional craftsmanship.
Authors: Zhijun Meng, Leiming Zhu, Gao Chen
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4415468523