A Deep Learning Approach to Intelligent Detection of Shedthin Tile Pathology in Suzhou Classical Gardens

A deep learning model, YOLO11-seg, achieved high accuracy in detecting four key pathologies—water stains, color aberration, surface scaling, and excessive gaps—in shedthin tiles from Suzhou Classical Gardens, with an overall accuracy of 74.38%. This approach significantly accelerates detection speed by an order of magnitude compared to traditional visual surveys, standardizes assessments across inspectors, and captures early-stage micro-pathologies often missed in manual inspections. The study provides a quantitative analysis of damage severity, offering critical references for scientific restoration strategies. This methodology sets a replicable template for preserving other fragile, repetition-rich historical materials in architectural heritage.

Chen, X., Wang, J., & Wang, S. (2025). A Deep Learning Approach to Intelligent Detection of Shedthin Tile Pathology in Suzhou Classical Gardens. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLVIII-M-9-2025, 239–246. https://doi.org/10.5194/isprs-archives-xlviii-m-9-2025-239-2025