Research on the identification and classification of deterioration in painting and calligraphy relics based on YOLO-seg

Painting and calligraphy relics are prone to various types of deterioration during long-term preservation, including Animal Damage, Missing Part, Crack, Folding Crease, and Stain. Traditional identification methods rely on manual observation, suffering from low efficiency and strong subjectivity. This study proposes a YOLO-seg-based instance segmentation method for deterioration identification and classification in high-resolution painting and calligraphy images. Five deterioration types are uniformly annotated as polygon instances. A training dataset is constructed through 1024 × 1024 patch cropping, target-centered supplementary sampling, and background negative sample retention. 2K full-image sliding-window inference with mask-NMS fusion achieves instance-level identification at full-image scale. Experimental results show an overall box mAP50 of 0.880 and a mask mAP50 of 0.875 across five categories. The method preserves the spatial distribution of deterioration regions and outputs instance masks, enabling quantitative preservation condition assessment from area, proportion, and connected component statistics.