AIGC-Driven Image Twin Generation and Evaluation for the Spatial Reconstruction of Ancient Academies: A Case Study of Classical Academy Imagery of Hunan Province

The digital reconstruction of ancient architecture and landscapes has become a prominent topic in current research. Existing studies predominantly focus on the three-dimensional reconstruction of archaeological sites or damaged historic buildings, while effective methods for the restoration or translation of architectural imagery recorded in ancient classical texts remain insufficient. Taking China’s distinctive educational architectural heritage—the ancient academy—as a case study, this paper proposes three Chinese AIGC methods, TI (Image to Image), TLTI (Image to Line Drawing to Image), and TPTI (Image to Prompts to Image). The methods use selected academy images from local gazetteers of Hunan Province, China, together with textual prompts to reconstruct the architecture and landscapes of ancient Chinese academies. The findings suggest that AIGC shows potential to support the expert-led reconstruction of landscapes and architectural imagery recorded in historical texts within a Chinese cultural context, enabling visualization of architectural developmental histories traditionally reliant on textual interpretation in the history of technology and art. This research offers a novel methodology for the integration of heritage science and the history of technology and art by activating cultural heritage in non-Western contexts through artificial intelligence and analyzing the evaluation challenges posed by AI models.