Digital Reconstruction of Hong'an Homespun Using AI and Semantic Differential Method

This study develops an innovative method for the attribution and visual reconstruction of hand-woven fabrics using artificial intelligence, employing Chinese Hong'an Homespun as a case study. The paper proposes a comprehensive algorithm integrating microscopic analysis, physical micro-model creation, and bimodal prompt engineering. The semantic differential method with a five-point scale was applied for objective evaluation of visual replica of historical fabrics. Comparative testing of AI models (Midjourney, ChatGPT, Qwen3, DouBaoAI, HailuoAI) revealed significant differences in their ability to reproduce characteristic features of hand weaving. The results demonstrate the superiority of detailed prompts with precise quantitative parameters and confirm the effectiveness of micro-models as visual anchors. The research establishes new standards in the digital documentation of cultural heritage and opens prospects for preserving traditional textile techniques. The most successful AI are Midjourney and ChatGPT have achieved an average score of 0.88 on the semantic scale, confirming the practical applicability of the method.