Anomaly Detection on MVTec-AD (I-AUROC, I-AP, I-F1-max)
99.8I-AUROCDDAD
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DDAD2024.06 | 99.8 | 99.5 | 97.9 | |
| DinomalyScaled training schedule=true2024.05 | 99.7 | 99.8 | 99.2 | |
| SimpleNet2024.06 | 99.6 | 99.6 | 98.9 | |
| Dinomaly2024.05 | 99.6 | 99.8 | 99 | |
| RAIDEvaluation Scenario=Full-shot, Paradigm=one-model-for-all-classes2026.02 | 99.4 | 99.8 | 98.7 | |
| GLADResolution=512x5122024.06 | 99.3 | 99.7 | 98.4 | |
| PatchCore2024.06 | 99.1 | 99.6 | 98.1 | |
| GLADResolution=256x2562024.06 | 99 | 99.7 | 98.2 | |
| Costfilter-ADEvaluation Scenario=Full-shot, Paradigm=one-model-for-all-classes2026.02 | 99 | 99.7 | 98.6 | |
| RD4AD2024.06 | 98.7 | 99.5 | 98 | |
| MambaADDesigned for MUAD=true2024.05 | 98.6 | 99.6 | 97.8 | |
| MambaADEvaluation Scenario=Full-shot, Paradigm=one-model-for-all-classes2026.02 | 98.6 | 99.6 | 97.8 | |
| OCR-GAN2024.06 | 98.3 | 98.1 | 95 | |
| ReContrastDesigned for MUAD=true2024.05 | 98.3 | 99.4 | 97.6 | |
| VITADDesigned for MUAD=true2024.05 | 98.3 | 99.4 | 97.3 | |
| ViTADEvaluation Scenario=Full-shot, Paradigm=one-model-for-all-classes2026.02 | 98.3 | 99.4 | 97.3 | |
| DRAEM2024.06 | 98 | 99 | 96.9 | |
| GLADEvaluation Scenario=Full-shot, Paradigm=one-model-for-all-classes2026.02 | 97.5 | 98.8 | 96.8 | |
| DiADDesigned for MUAD=true2024.05 | 97.2 | 99 | 96.5 | |
| DiADEvaluation Scenario=Full-shot, Paradigm=one-model-for-all-classes2026.02 | 97.2 | 99 | 96.5 | |
| AnomalyDINOEvaluation Scenario=Full-shot, Paradigm=one-model-for-all-classes2026.02 | 96.8 | 98.6 | 97.1 | |
| UniADDesigned for MUAD=true2024.05 | 96.5 | 98.8 | 96.2 | |
| UniADEvaluation Scenario=Full-shot, Paradigm=one-model-for-all-classes2026.02 | 96.5 | 98.8 | 96.2 | |
| PatchCoreEvaluation Scenario=Full-shot, Paradigm=one-model-for-all-classes2026.02 | 96.4 | — | — | |
| SimpleNet2024.05 | 95.3 | 98.4 | 95.8 | |
| SimpleNetEvaluation Scenario=Full-shot, Paradigm=one-model-for-all-classes2026.02 | 95.3 | 98.4 | 95.8 | |
| RD4AD2024.05 | 94.6 | 96.5 | 95.2 | |
| Bayes-PFLZero-shot=true, GPU cost (GB)=3.3, Time (ms)=388.52025.03 | 92.3 | 96.7 | — | |
| AA-CLIPAvailable training shots=642025.03 | 92 | — | — | |
| AdaCLIPZero-shot=true, GPU cost (GB)=3.3, Time (ms)=183.42025.03 | 92 | 96.4 | — | |
| WinCLIPAvailable training shots=N/A2025.03 | 91.8 | — | — | |
| WinCLIPZero-shot=true, GPU cost (GB)=2.0, Time (ms)=840.02025.03 | 91.8 | 95.1 | — | |
| AnomalyCLIPZero-shot=true, GPU cost (GB)=2.7, Time (ms)=131.22025.03 | 91.5 | 96.2 | — | |
| AnomalyCLIPAvailable training shots=full2025.03 | 90.9 | — | — | |
| AA-CLIPAvailable training shots=full2025.03 | 90.5 | — | — | |
| AdaCLIPAvailable training shots=full2025.03 | 90 | — | — | |
| CLIP-ADZero-shot=true, GPU cost (GB)=3.4, Time (ms)=115.22025.03 | 89.8 | 95.3 | — | |
| AA-CLIPAvailable training shots=162025.03 | 89.7 | — | — | |
| DeSTSeg2024.05 | 89.2 | 95.5 | 91.6 | |
| MVFA-ADAvailable training shots=full2025.03 | 86.6 | — | — | |
| CLIP & VANDAvailable training shots=N/A2025.03 | 86.1 | — | — | |
| APRIL-GANZero-shot=true, GPU cost (GB)=3.3, Time (ms)=105.02025.03 | 86.1 | 93.5 | — | |
| AA-CLIPAvailable training shots=22025.03 | 85.9 | — | — |