Anomaly Detection on MVTec-AD (test) (I-AUROC, P-AUROC, P-AUPRO)
100I-AUROCCCAD(C)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CCAD(C)Class=Bottle, Training Paradigm=Multi-class based2025.12 | 100 | 98.6 | — | |
| CCAD(V)Class=Grid, Training Paradigm=Single class based2025.12 | 100 | 99.3 | — | |
| DiADClass=Leather, Training Paradigm=Multi-class based2025.12 | 100 | 99.1 | — | |
| CCAD(C)Class=Leather, Training Paradigm=Multi-class based2025.12 | 100 | 97.9 | — | |
| CCAD(F)Class=Leather, Training Paradigm=Multi-class based2025.12 | 100 | 99 | — | |
| CCAD(V)Class=Metal Nut, Training Paradigm=Single class based2025.12 | 100 | 98.5 | — | |
| DDADClass=Tile, Training Paradigm=Single class based2025.12 | 100 | 97.9 | — | |
| CCAD(V)Class=Tile, Training Paradigm=Single class based2025.12 | 100 | 97.9 | — | |
| CCAD(V)Class=Toothbrush, Training Paradigm=Single class based2025.12 | 100 | 98.4 | — | |
| DDADClass=Grid, Training Paradigm=Single class based2025.12 | 99.9 | 99.3 | — | |
| PMSR2026.06 | 99.9 | — | — | |
| GLASS2026.06 | 99.9 | — | — | |
| PAE2026.06 | 99.9 | — | — | |
| DDAD-10Number of denoising steps=10, Inference time (s)=0.382023.05 | 99.8 | 98.1 | 92.4 | |
| CCAD(F)Class=Bottle, Training Paradigm=Multi-class based2025.12 | 99.8 | 98.6 | — | |
| CCAD(V)Class=Wood, Training Paradigm=Single class based2025.12 | 99.8 | 93.1 | — | |
| PBAS2026.06 | 99.8 | — | — | |
| DinomalySetting=class-separated2024.05 | 99.7 | 99.9 | 95 | |
| DDAD-S-10Number of denoising steps=10, Variant=Compressed (S), Inference time (s)=0.342023.05 | 99.7 | 97.9 | 91.3 | |
| DDAD-25Number of denoising steps=25, Inference time (s)=0.902023.05 | 99.7 | 97.9 | 91 | |
| RealNet2026.06 | 99.7 | — | — | |
| DinomalySetting=MUAD2024.05 | 99.6 | 98.4 | 94.8 | |
| SimpleNetSetting=class-separated2024.05 | 99.6 | 98.1 | 90 | |
| DiADClass=Bottle, Training Paradigm=Multi-class based2025.12 | 99.6 | 98.4 | — | |
| SimpleNet2026.06 | 99.6 | — | — | |
| PatchCoreClass=Bottle, Training Paradigm=Single class based2025.12 | 99.5 | 97.9 | — | |
| CCAD(F)Class=Grid, Training Paradigm=Multi-class based2025.12 | 99.5 | 99 | — | |
| PatchCoreClass=Zipper, Training Paradigm=Single class based2025.12 | 99.5 | 98.7 | — | |
| DiADClass=Toothbrush, Training Paradigm=Multi-class based2025.12 | 99.4 | 99 | — | |
| DDAD-5Number of denoising steps=5, Inference time (s)=0.212023.05 | 99.3 | 97.5 | 91.2 | |
| CCAD(C)Class=Carpet, Training Paradigm=Multi-class based2025.12 | 99.3 | 98.7 | — | |
| CCAD(F)Class=Carpet, Training Paradigm=Multi-class based2025.12 | 99.2 | 98.7 | — | |
| CCAD(C)Class=Transistor, Training Paradigm=Multi-class based2025.12 | 99.2 | 95.5 | — | |
| DDADClass=Wood, Training Paradigm=Single class based2025.12 | 99.2 | 94.4 | — | |
| FODSetting=Separate2026.03 | 99.2 | 98.3 | — | |
| FSRSetting=Separate, tau=Optimal2026.03 | 99.2 | 98.4 | — | |
| PatchCore2026.06 | 99.2 | — | — | |
| PatchCoreSetting=class-separated2024.05 | 99.1 | 98.1 | 93.5 | |
| PatchCoreSampling percentage=10%, Inference time (s)=0.222023.05 | 99.1 | 98.1 | 93.5 | |
| DDADClass=Metal Nut, Training Paradigm=Single class based2025.12 | 99.1 | 98.2 | — | |
| FSRSetting=Separate, tau=0.12026.03 | 99.1 | 98.4 | — | |
| PatchCoreSampling percentage=1%, Inference time (s)=0.172023.05 | 99 | 98.1 | 93.5 | |
| DiADClass=Carpet, Training Paradigm=Multi-class based2025.12 | 99 | 98.2 | — | |
| PatchCoreClass=Leather, Training Paradigm=Single class based2025.12 | 98.8 | 99.4 | — | |
| DiADClass=Metal Nut, Training Paradigm=Multi-class based2025.12 | 98.7 | 97.8 | — | |
| CCAD(C)Class=Wood, Training Paradigm=Multi-class based2025.12 | 98.7 | 93 | — | |
| CCAD(F)Class=Wood, Training Paradigm=Multi-class based2025.12 | 98.7 | 93.2 | — | |
| CCAD(V)Class=Zipper, Training Paradigm=Single class based2025.12 | 98.7 | 97.5 | — | |
| RD4ADSetting=class-separated2024.05 | 98.5 | 97.8 | 93.9 | |
| CCAD(F)Class=Tile, Training Paradigm=Multi-class based2025.12 | 98.3 | 92.9 | — | |
| FSRSetting=Unified, tau=Optimal2026.03 | 98.3 | 98 | — | |
| DiADClass=Wood, Training Paradigm=Multi-class based2025.12 | 98.2 | 91.8 | — | |
| DDADClass=Toothbrush, Training Paradigm=Single class based2025.12 | 98.1 | 98.5 | — | |
| DRAEM2026.06 | 98 | — | — | |
| DDADClass=Zipper, Training Paradigm=Single class based2025.12 | 97.9 | 95.5 | — | |
| CCAD(F)Class=Metal Nut, Training Paradigm=Multi-class based2025.12 | 97.8 | 97.1 | — | |
| CCAD(F)Class=Hazelnut, Training Paradigm=Multi-class based2025.12 | 97.7 | 98.1 | — | |
| CCAD(V)Class=Bottle, Training Paradigm=Single class based2025.12 | 97.5 | 93.9 | — | |
| CCAD(F)Class=Toothbrush, Training Paradigm=Multi-class based2025.12 | 97.5 | 96.2 | — | |
| SPADEClass=Bottle, Training Paradigm=Single class based2025.12 | 97.2 | 56.2 | — | |
| CCAD(C)Class=Hazelnut, Training Paradigm=Multi-class based2025.12 | 97.2 | 97.2 | — | |
| DiADSetting=Unified2026.03 | 97.2 | 96.8 | — | |
| DiADClass=Hazelnut, Training Paradigm=Multi-class based2025.12 | 97.1 | 97.3 | — | |
| CCAD(C)Class=Metal Nut, Training Paradigm=Multi-class based2025.12 | 97.1 | 97.8 | — | |
| CCAD(F)Class=Transistor, Training Paradigm=Multi-class based2025.12 | 97.1 | 92.1 | — | |
| FSRSetting=Mean, tau=Optimal2026.03 | 97 | 98.1 | — | |
| PatchCoreClass=Wood, Training Paradigm=Single class based2025.12 | 96.8 | 94.1 | — | |
| CCAD(V)Class=mean, Training Paradigm=Single class based2025.12 | 96.8 | 96.5 | — | |
| CCAD(V)Class=Screw, Training Paradigm=Single class based2025.12 | 96.7 | 99.2 | — | |
| CCAD(C)Class=Tile, Training Paradigm=Multi-class based2025.12 | 96.6 | 92.3 | — | |
| SPADEClass=Zipper, Training Paradigm=Single class based2025.12 | 96.6 | 41.5 | — | |
| FSRSetting=Unified, tau=0.12026.03 | 96.6 | 97.9 | — | |
| CCAD(V)Class=Pill, Training Paradigm=Single class based2025.12 | 96.5 | 97.4 | — | |
| SPADEClass=Tile, Training Paradigm=Single class based2025.12 | 96.5 | 63.2 | — | |
| DiADClass=Tile, Training Paradigm=Multi-class based2025.12 | 96.5 | 92.6 | — | |
| DDADClass=Screw, Training Paradigm=Single class based2025.12 | 96.4 | 99.2 | — | |
| FSRSetting=Mean, tau=0.12026.03 | 96.4 | 98 | — | |
| CCAD(C)Class=Toothbrush, Training Paradigm=Multi-class based2025.12 | 96.1 | 98.9 | — | |
| CCAD(F)Class=mean, Training Paradigm=Multi-class based2025.12 | 96.1 | 96.2 | — | |
| CCAD(V)Class=Transistor, Training Paradigm=Single class based2025.12 | 96 | 89.4 | — | |
| CCAD(V)Class=Cable, Training Paradigm=Single class based2025.12 | 95.9 | 96 | — | |
| SPADEClass=Wood, Training Paradigm=Single class based2025.12 | 95.9 | 64.4 | — | |
| DDADClass=Pill, Training Paradigm=Single class based2025.12 | 95.7 | 97.2 | — | |
| FastReconSetting=Separate2026.03 | 95.7 | 97.8 | — | |
| PaDiMInference time (s)=0.192023.05 | 95.4 | 97.5 | 92.1 | |
| SPADEClass=Leather, Training Paradigm=Single class based2025.12 | 95.4 | 61.5 | — | |
| CCAD(F)Class=Zipper, Training Paradigm=Multi-class based2025.12 | 95.4 | 95.9 | — | |
| CCAD(C)Class=mean, Training Paradigm=Multi-class based2025.12 | 95.3 | 95.9 | — | |
| FastReconSetting=Few-shot-82026.03 | 95.2 | 97.3 | — | |
| CutPaste2026.06 | 95.2 | — | — | |
| FODSetting=Mean2026.03 | 95.1 | 97 | — | |
| DiADClass=mean, Training Paradigm=Multi-class based2025.12 | 95 | 95.4 | — | |
| DiADClass=Grid, Training Paradigm=Multi-class based2025.12 | 94.9 | 94.7 | — | |
| DDADClass=Cable, Training Paradigm=Single class based2025.12 | 94.7 | 96.5 | — | |
| DDADClass=Transistor, Training Paradigm=Single class based2025.12 | 94.7 | 88 | — | |
| CCAD(C)Class=Grid, Training Paradigm=Multi-class based2025.12 | 94.5 | 93 | — | |
| DiADClass=Transistor, Training Paradigm=Multi-class based2025.12 | 94.5 | 89.9 | — | |
| CCAD(V)Class=Capsule, Training Paradigm=Single class based2025.12 | 94.3 | 96.1 | — | |
| DDADClass=mean, Training Paradigm=Single class based2025.12 | 94.3 | 95.6 | — | |
| CCAD(C)Class=Pill, Training Paradigm=Multi-class based2025.12 | 94.2 | 96.4 | — |