Application of computer vision technology in surface damage detection and analysis of shedthin tiles in China: a case study of the classical gardens of Suzhou

A machine learning-based approach using the YOLOv4 model demonstrates high accuracy in detecting and classifying surface damage on shedthin tiles, a critical material in China's UNESCO-listed Classical Gardens of Suzhou. The model achieved comprehensive accuracy rates of 90.20%, with specific accuracies of 85.89% for water stains, 93.29% for surface scaling, 87.37% for color aberration, and 96.15% for gaps—exceeding basic testing requirements. This innovation overcomes traditional assessment limitations by automating damage detection, reducing time and labor costs while maintaining reliability in complex environments.

Yan, L., Chen, Y., Zheng, L., & Zhang, Y. (2024). Application of computer vision technology in surface damage detection and analysis of shedthin tiles in China: a case study of the classical gardens of Suzhou. Heritage Science, 12(1). https://doi.org/10.1186/s40494-024-01185-6