State-of-the-Art Fails in the Art of Damage Detection
This addresses a critical challenge in cultural heritage preservation by highlighting limitations in current damage detection methods, though it is incremental as it focuses on dataset creation and evaluation without proposing a new solution.
The paper tackled the problem of accurately detecting and classifying damage in analogue media for cultural heritage preservation, showing that existing machine learning models fail to predict damage locations even after supervised training, and introduced DamBench, a dataset with over 11,000 annotations covering 15 damage types.
Accurately detecting and classifying damage in analogue media such as paintings, photographs, textiles, mosaics, and frescoes is essential for cultural heritage preservation. While machine learning models excel in correcting global degradation if the damage operator is known a priori, we show that they fail to predict where the damage is even after supervised training; thus, reliable damage detection remains a challenge. We introduce DamBench, a dataset for damage detection in diverse analogue media, with over 11,000 annotations covering 15 damage types across various subjects and media. We evaluate CNN, Transformer, and text-guided diffusion segmentation models, revealing their limitations in generalising across media types.