Digital three-dimensional reconstruction of built heritage through algorithmic assessment and media-driven semantic mining analysis

Abstract This study investigates how digital 3D reconstruction of built heritage can be evaluated not only in terms of morphological performance but also in terms of media-driven cultural perception. Using the Three Shadow Pagoda as a case study, we pursued two objectives: to analyze public cognition from 2942 TikTok comments and to compare five image-to-3D algorithms using a multi-metric framework. Results show that engagement and cultural cognition dominated public responses, while “buildings and structure” and “culture and history” were the strongest semantic clusters. Hunyuan3D-2 achieved the highest overall model quality assessment score, followed by Rodin and SPAR3D, and also produced the highest perceptual similarity score. The study’s novelty lies in integrating social media text mining, relative morphological evaluation, and an interpretive media memory model to reposition digital heritage reconstruction from technical restoration toward cultural representation.