An Intelligent Design Framework for Regional Cultural Creative Products Using Generative AI: A Jilin Ice-Snow Case Study
Year: 2026
Type: conference-paper
Abstract: The design of regional cultural creative products faces a fundamental conflict between unstructured cultural source materials and the structured visual requirements of modern product design pipelines. State-of-the-art text-to-image diffusion models accept only flat natural-language prompts and therefore cannot capture the hierarchical semantics of regional culture, producing designs that are visually plausible but culturally shallow. We propose a three-step intelligent design workflow that transforms unstructured cultural corpora to an 82-dimensional conditioning vector \({\mathrm{C\ }} = {\mathrm{\ }}[ {{\mathrm{V}};{\mathrm{\ P}};{\mathrm{\ S}}} ]\) involving visual elements, color schemes, and symbolism annotations. The conditioning vector is plugged into a LoRA-tuned Stable Diffusion pipeline, and output candidates are evaluated through closed-loop evaluation using CLIP and NIMA metrics. Experiments in the Jilin ice-and-snow cultural region show a 45% increase in cultural relevance compared to the baseline prompt-only approach and accelerate design completion by about 28 seconds per batch, representing a five-order-of-magnitude boost in productivity over conventional manual design procedures.
Author: Yan Zhang
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W7213450481