Cross-cultural insights into traditional Jiangnan gardens of China and Japanese gardens through algorithm-enhanced comparative analysis
摘要
Traditional Jiangnan gardens in China and traditional Japanese gardens are distinguished not only by their unique aesthetic features but also by their deep philosophical underpinnings. Leveraging the advanced YOLOv8 deep learning model enhanced with a Self-Attention (SA) module, this study introduces a novel approach to distinguishing these garden styles. The enhanced YOLOv8-SA model efficiently processes a custom dataset of 3606 high-resolution images, achieving a precision of 93.9%, a recall of 94.7%, and an mAP@50 of 98.6%. This study underscores the role of artificial intelligence in meticulously identifying and cataloging the nuanced elements of traditional gardens, thereby facilitating their conservation as vital cultural and historical heritages. By harnessing advanced computational techniques to analyze and document the intricate aesthetics of Eastern garden art, this research contributes to heritage science goals—ensuring that these living heritage sites are preserved and appreciated for future generations.