Anomaly Severity Classification is a novel task that extends traditional anomaly detection by distinguishing between negligible anomalies and critical defects. While existing benchmarks focus on identifying whether an input deviates from normality, this task introduces a deployment-oriented perspective—evaluating the actionability of detected anomalies. For example, cosmetic deviations or design shifts may be acceptable (anomalies), while structural damages require intervention (defects). This task supports more informed and automated decision-making in industrial and quality control settings. It was introduced in the VELM paper, alongside a simulation benchmark derived from the MVTec-AC dataset.
데이터셋
출처에 같은 태스크 이름으로 연결된 데이터셋입니다. 동명 태스크가 포함될 수 있으므로, 비교할 때는 평가 결과의 조건도 확인하세요.
MVTec-AC is a curated refinement of the widely-used MVTec-AD dataset, specifically designed for anomaly classification—distinguishing between different types of anomalies rather than merely detecting if an image is anomalous. While MVTec-AD focuses on binary detection and suffers from mislabeled or ambiguous samples, MVTec-AC introduces manually corrected labels and reorganized anomaly categories to enable robust multi-class evaluation. Key improvements include the correction of 36 misclassified samples, merging of 4 overlapping classes, removal of 4 ambiguous ‘combined’ classes, and exclusion of the toothbrush category, which contains only a single trivial anomaly type. These changes support consistent, fine-grained assessment of classification models in industrial visual inspection contexts.