An improved Stingray model for applying Yi ethnic clothing patterns to daily ceramics for intangible cultural heritage

In the design of ethnic cultural products, generative AI often leads to the loss of cultural characteristics and stylistic deviations due to the lack of systematic methodological guidance. This paper proposes a generative AI style transfer framework based on an improved Stingray model. It constructs a “pattern gene bank” to extract cultural features, integrates Kano model user demand analysis to prioritize design elements and clarify objectives, and combines Stable Diffusion's controllable generation with TOPSIS multi-objective evaluation. This establishes a structured methodology from cultural decoding to design output. Using the innovative adaptation of Yi ethnic clothing patterns onto daily ceramic tableware as a case study, this paper employs generative tools to achieve design innovations that balance cultural accuracy with modern aesthetics. The feasibility is validated via the TOPSIS method, where the optimal solution's relative proximity (Ci = 0.742) significantly outperforms existing market solutions. The findings demonstrate that the enhanced Stingray model effectively integrates multiple design methodologies, addressing feature loss and feasibility limitations inherent in single-model approaches during cultural inheritance and style transfer. Concurrently, system-guided generative tools substantially boost design efficiency and streamline solution refinement, significantly shortening design cycles while enhancing proposal diversity. This study provides a systematic methodological framework for the application of generative artificial intelligence in the design of ethnic cultural products and opens up new avenues for the intelligent transformation of the traditional ceramics industry.