Research on the Renovation of Industrial Architectural Remains in Rural Areas of Southern Anhui Province Based on Generative Artificial Intelligence and Human Perception Experiments

Southern Anhui Province retains a significant number of rural industrial architectural remains, which face widespread abandonment and deterioration due to functional decline and insufficient preservation. Traditional design methods suffer from inefficiency and a lack of public perception validation, struggling to balance historical authenticity with modern functionality. This study innovatively integrates generative artificial intelligence and human perception experiments to propose an efficient renovation approach. By constructing a database of industrial architectural remains in Southern Anhui and combining the Stable Diffusion model with lightweight finetuning (LoRA), we generated renovation schemes that harmonize regional style and functional requirements. Experiments demonstrated that adjusting model parameters and textual prompts enables spatial renovation while preserving architectural heritage styles. Further validation through eye-tracking and Semantic Differential (SD) methods revealed a significant increase in heatmap focus concentration for updated images. SD scores indicated high evaluations for renovation effectiveness, though optimization of cultural characteristics remains necessary.