Optimal design of ceramic form combining stable diffusion model and GRU-Attention

As a treasure of intangible cultural heritage, Chinese ceramic art plays an important role in the world cultural heritage. However, the advancement of industrialization and the aging of craftsmen have led to difficulties in the inheritance and innovation of porcelain making techniques. The aim of this study is to integrate the stable diffusion model and GRU-Attention to accurately capture the Kansei needs of customers and optimize the shape design for fashionable ceramics. Firstly, on the basis of collecting a large number of ceramic images on the market, the stable diffusion model is used to extensively learn the morphological characteristics and distribution rules of ceramics, and then 150 innovative forms that have not been seen on the market are generated. Secondly, factor analysis is used to select Kansei words that can best represent customers’ aesthetic preferences and they are matched with 150 ceramic samples to construct Kansei evaluation matrix . Then, GRU-Attention establishes the mapping relationship between customers’ Kansei demands and ceramic morphological characteristics, thus the best form combination with the highest value of each Kansei evaluation is determined. Compared with other neural networks such as CNN , LSTM , RBF and BPNN, GRU-Attention can better capture long-term dependencies in data, thus improving the accuracy of prediction and classification. Finally, according to the optimal form combination parameters, the fashion designers collaboratively create the ceramic shape design that meets the customers’ Kansei needs. The model proposed in this paper automatically generates creative and diverse ceramic forms, which is helpful for breaking through the limitations of traditional design and provides novel concepts for the innovative development of excellent traditional Chinese culture.