This study takes the entrances of three modern industrial heritage sites, including the Nanjing Jinling Arsenal, as its research subjects. By integrating AIGC image restoration technology, eye tracking analysis (ET), and the semantic difference method (SD), and combining objective data with subjective evaluations, this study explores the visual perception effects of the entrances of modern industrial heritage sites in China. First, historical photographs of the gates are restored using the Stable Diffusion tool. Next, eye movement data are obtained from the gate through a 45-subject eye-movement experiment. Finally, subjective perception is quantified using semantic differentials, and three principal components, including esthetic design, were extracted. The results indicate that the area of the region of interest is significantly positively correlated with certain eye movement metrics; the lintel has higher visual priority owing to its information density advantage, and objective and subjective data are highly correlated, with eye-tracking heatmaps and questionnaire scores showing high consistency. The innovation of this study lies in combining the AOI area standardization method to eliminate scale interference and thereby revealing that both physical scale and semantic features influence the visual appeal of industrial heritage. This study provides references for the core area maintenance, material expression, and visual optimization of industrial heritage gates.
Li et al. (2026) studied this question.