Xiangyunsha Defect Dataset and a Lightweight Detection Model

This dataset contains 1,831 annotated images of Xiangyunsha (Gambiered Guangdong silk) surface defects, collected through participatory fieldwork in three heritage-certified workshops in Foshan, China. The dataset includes five defect categories: crackle anomaly, stain patch, crease shadow, edge discoloration, and spot damage. All images were annotated using Labelme, with bounding-box labels for each defect instance. This dataset supports the research presented in the paper "Constructing a Xiangyunsha Defect Dataset and a Lightweight Detection Model for Sustainability," submitted to Sustainability. The dataset is released under the CC BY 4.0 license.