A study on the path of Lingnan architectural style in China based on AI generation technology

Abstract

This paper explores the application of AI-generated content in architectural design, focusing on the Lingnan architectural style. Using Stable Diffusion and LoRA fine-tuning models, this study simulates and generates typical design features of Lingnan architecture by training on architectural images. Conditional control methods, such as ControlNet, are incorporated to enhance spatial structure recognition and architectural detail, ensuring precise outputs. Additionally, the study examines a hybrid generation approach, blending traditional Lingnan and modern architectural styles to evaluate potential style transitions and innovations. Findings suggest that AI generation technology effectively captures Lingnan architectural details while fostering style integration and evolution. This research provides a valuable technical and theoretical foundation for the digital preservation of Lingnan architecture and contemporary design.