Abstract. Ancient Chinese architectural polychromy embodies profound cultural significance and aesthetic value. However, it faces two main challenges: the evolving demands of contemporary design and the discontinuity in traditional craftsmanship transmission. This study focuses on polychrome patterns of the Song Dynasty and proposes a framework for their digital preservation and innovative redesign based on Artificial Intelligence-Generated Content (AIGC) technologies. Grounded in canonical texts such as the Yingzao Fashi (营 造法式， Treatise on Architectural Methods), the research systematically analyzes the formal, chromatic, and symbolic characteristics—termed “form–color–meaning”—of Song Dynasty architectural polychrome decoration. A representative dataset was constructed, comprising typical pattern types such as Huawenjin ( 花纹锦, Floral Brocade), Suowenjin ( 琐纹锦, Interlocking Brocade), Qushuijin ( 曲水锦, Meandering-Stream Brocade), Jingdijin ( 净地锦, Plain-Ground Brocade). By combining Stable Diffusion v3.5 with the LoRA fine-tuning mechanism (including a human–machine collaborative feedback loop), the framework integrates multidimensional evaluations of cultural semantics and aesthetic quality into its iterative optimization cycles, the approach achieves high-fidelity structural reconstruction and deep semantic expression. Experimental results demonstrate that the model excels in structural reconstruction and stylistic emulation. The generated patterns exhibit robust cultural adaptability and innovative design capacity across applications such as virtual exhibitions, cultural – creative products, and interactive education. The developed "Songfeng Caiyun" ( 宋风彩韵, meaning 'Song-style Aesthetic Chroma') intelligent platform further validates the proposed framework’s feasibility for digital conservation, creative redesign, and cross-media dissemination of traditional Chinese ornamental art.
Liu et al. (Tue,) studied this question.