Dinomaly2: A Unified Framework for Unsupervised Image Anomaly Detection

Benchmark Model Rank Results
multi-class-anomaly-detection-on-mvtec-adDinomaly2-L#1Detection AUROC: 99.9Segmentation AUROC: 98.6Segmentation AP: 70.3
multi-class-anomaly-detection-on-real-iad-unified-multi-classDinomaly2-L#1Detection AUROC: 92.1Segmentation AP: 48.4Detection F1-max: 83.4…
multi-class-anomaly-detection-on-visa-unified-multi-classDinomaly2-L#1Detection AUROC: 99.3Segmentation AUROC: 99.2Detection F1-max: 97.2…
rgb-3d-anomaly-detection-and-segmentation-on-mvtec-3d-ad-rgb-3d-fusionDinomaly2-L#2Detection AUROC: 97.4Segmentation AUPRO@30% FPR: 98.6…
anomaly-detection-on-mvtec-adDinomaly2-B#8Detection AUROC: 99.8Segmentation AUROC: 98.4