OOD Detection on MVTec Pill (Dissimilar)
86.15AUROCKNN
Evaluation results
| Method | Links | |
|---|---|---|
| KNNBackbone=ResNet18, Method Category=Feature-based, Hyperparameter Setting=best-performing2026.02 | 86.15 | |
| KNNBackbone=VGG16, Method category=Feature-based, Hyperparameter selection=best-performing2026.02 | 82.82 | |
| DICEBackbone=ResNet18, Method Category=Confidence-based, Hyperparameter Setting=best-performing2026.02 | 82.56 | |
| NMDBackbone=ResNet18, Method Category=Feature-based, Hyperparameter Setting=best-performing2026.02 | 82.31 | |
| CoPBackbone=ResNet18, Method Category=Feature-based, Hyperparameter Setting=best-performing2026.02 | 81.15 | |
| ODINBackbone=ResNet18, Method Category=Confidence-based, Hyperparameter Setting=best-performing2026.02 | 80.77 | |
| ViMBackbone=ResNet18, Method Category=Confidence-based, Hyperparameter Setting=best-performing2026.02 | 80.77 | |
| ASHBackbone=ResNet18, Method Category=Confidence-based, Hyperparameter Setting=best-performing2026.02 | 80.64 | |
| WeiPerBackbone=ResNet18, Method Category=Confidence-based, Hyperparameter Setting=best-performing2026.02 | 80.29 | |
| SHEBackbone=ResNet18, Method Category=Confidence-based, Hyperparameter Setting=best-performing2026.02 | 80.02 | |
| ReActBackbone=ResNet18, Method Category=Confidence-based, Hyperparameter Setting=best-performing2026.02 | 80 | |
| Deep EnsembleBackbone=ResNet18, Method Category=Confidence-based, Hyperparameter Setting=best-performing2026.02 | 79.51 | |
| GradNormBackbone=ResNet18, Method Category=Confidence-based, Hyperparameter Setting=best-performing2026.02 | 79.07 | |
| MC-DropoutBackbone=ResNet18, Method Category=Confidence-based, Hyperparameter Setting=best-performing2026.02 | 79.04 | |
| GradOrthBackbone=ResNet18, Method Category=Confidence-based, Hyperparameter Setting=best-performing2026.02 | 78.59 | |
| MCPBackbone=ResNet18, Method Category=Confidence-based, Hyperparameter Setting=best-performing2026.02 | 78.33 | |
| ViMBackbone=VGG16, Method category=Confidence-based, Hyperparameter selection=best-performing2026.02 | 76.18 | |
| ResidualBackbone=VGG16, Method category=Feature-based, Hyperparameter selection=best-performing2026.02 | 76.09 | |
| KDE (Gaussian)Backbone=ResNet18, Method Category=Feature-based, Hyperparameter Setting=best-performing2026.02 | 73.59 | |
| NuSABackbone=ResNet18, Method Category=Feature-based, Hyperparameter Setting=best-performing2026.02 | 73.21 | |
| CoRPBackbone=ResNet18, Method Category=Feature-based, Hyperparameter Setting=best-performing2026.02 | 71.41 | |
| MBMBackbone=ResNet18, Method Category=Feature-based, Hyperparameter Setting=best-performing2026.02 | 71.36 | |
| KNNBackbone=ViT-B/32, Method Type=Feature-based2026.02 | 71.28 | |
| GRAMBackbone=ResNet18, Method Category=Feature-based, Hyperparameter Setting=best-performing2026.02 | 69.83 | |
| MahalanobisBackbone=VGG16, Method category=Feature-based, Hyperparameter selection=best-performing2026.02 | 69.78 | |
| MBMBackbone=VGG16, Method category=Feature-based, Hyperparameter selection=best-performing2026.02 | 69.58 | |
| CoPBackbone=VGG16, Method category=Feature-based, Hyperparameter selection=best-performing2026.02 | 69.07 | |
| ResidualBackbone=ResNet18, Method Category=Feature-based, Hyperparameter Setting=best-performing2026.02 | 68.97 | |
| XOOD-MBackbone=VGG16, Method category=Feature-based, Hyperparameter selection=best-performing2026.02 | 68.84 | |
| XOOD-MBackbone=ViT-B/32, Method Type=Feature-based2026.02 | 68.73 | |
| MahalanobisBackbone=ResNet18, Method Category=Feature-based, Hyperparameter Setting=best-performing2026.02 | 68.72 | |
| LOFBackbone=VGG16, Method category=Feature-based, Hyperparameter selection=best-performing2026.02 | 68.72 | |
| NACBackbone=ResNet18, Method Category=Feature-based, Hyperparameter Setting=best-performing2026.02 | 67.88 | |
| CoPBackbone=ViT-B/32, Method Type=Feature-based2026.02 | 66.92 | |
| XOOD-MBackbone=ResNet18, Method Category=Feature-based, Hyperparameter Setting=best-performing2026.02 | 66.67 | |
| PCXBackbone=VGG16, Method category=Feature-based, Hyperparameter selection=best-performing2026.02 | 66.29 | |
| MBMBackbone=ViT-B/32, Method Type=Feature-based2026.02 | 66.27 | |
| CoRPBackbone=VGG16, Method category=Feature-based, Hyperparameter selection=best-performing2026.02 | 65.77 | |
| TAPUUDBackbone=ViT-B/32, Method Type=Feature-based2026.02 | 65.64 | |
| MahalanobisBackbone=ViT-B/32, Method Type=Feature-based2026.02 | 65.22 | |
| NMDBackbone=ViT-B/32, Method Type=Feature-based2026.02 | 64.87 | |
| Deep EnsembleBackbone=ViT-B/32, Method Type=Confidence-based2026.02 | 64.75 | |
| CoRPBackbone=ViT-B/32, Method Type=Feature-based2026.02 | 64.38 | |
| ODINBackbone=ViT-B/32, Method Type=Confidence-based2026.02 | 63.91 | |
| ResidualBackbone=ViT-B/32, Method Type=Feature-based2026.02 | 63.72 | |
| PCXBackbone=ViT-B/32, Method Type=Feature-based2026.02 | 63.15 | |
| FeatureNormBackbone=ViT-B/32, Method Type=Feature-based2026.02 | 62.49 | |
| GradNormBackbone=ViT-B/32, Method Type=Confidence-based2026.02 | 62.46 | |
| LOFBackbone=ResNet18, Method Category=Feature-based, Hyperparameter Setting=best-performing2026.02 | 62.18 | |
| WeiPerBackbone=ViT-B/32, Method Type=Confidence-based2026.02 | 62.18 | |
| Deep EnsembleBackbone=VGG16, Method category=Confidence-based, Hyperparameter selection=best-performing2026.02 | 60.62 | |
| ASHBackbone=ViT-B/32, Method Type=Confidence-based2026.02 | 60.51 | |
| GRAMBackbone=VGG16, Method category=Feature-based, Hyperparameter selection=best-performing2026.02 | 60.19 | |
| GradOrthBackbone=VGG16, Method category=Confidence-based, Hyperparameter selection=best-performing2026.02 | 60.18 | |
| ASHBackbone=VGG16, Method category=Confidence-based, Hyperparameter selection=best-performing2026.02 | 60.13 | |
| NuSABackbone=ViT-B/32, Method Type=Feature-based2026.02 | 59.1 | |
| SHEBackbone=ViT-B/32, Method Type=Confidence-based2026.02 | 59.04 | |
| GradOrthBackbone=ViT-B/32, Method Type=Confidence-based2026.02 | 59.02 | |
| NANBackbone=ViT-B/32, Method Type=Feature-based2026.02 | 58.85 | |
| TAPUUDBackbone=VGG16, Method category=Feature-based, Hyperparameter selection=best-performing2026.02 | 58.46 | |
| GradNormBackbone=VGG16, Method category=Confidence-based, Hyperparameter selection=best-performing2026.02 | 58.42 | |
| NMDBackbone=VGG16, Method category=Feature-based, Hyperparameter selection=best-performing2026.02 | 57.95 | |
| LOFBackbone=ViT-B/32, Method Type=Feature-based2026.02 | 57.69 | |
| NANBackbone=ResNet18, Method Category=Feature-based, Hyperparameter Setting=best-performing2026.02 | 56.92 | |
| FeatureNormBackbone=ResNet18, Method Category=Feature-based, Hyperparameter Setting=best-performing2026.02 | 56.41 | |
| PCXBackbone=ResNet18, Method Category=Feature-based, Hyperparameter Setting=best-performing2026.02 | 56.41 | |
| MC-DropoutBackbone=VGG16, Method category=Confidence-based, Hyperparameter selection=best-performing2026.02 | 55.31 | |
| MCPBackbone=ViT-B/32, Method Type=Confidence-based2026.02 | 55.26 | |
| MC-DropoutBackbone=ViT-B/32, Method Type=Confidence-based2026.02 | 54.96 | |
| GRAMBackbone=ViT-B/32, Method Type=Feature-based2026.02 | 54.83 | |
| KDE (Gaussian)Backbone=VGG16, Method category=Feature-based, Hyperparameter selection=best-performing2026.02 | 54.74 | |
| SHEBackbone=VGG16, Method category=Confidence-based, Hyperparameter selection=best-performing2026.02 | 54.1 | |
| ODINBackbone=VGG16, Method category=Confidence-based, Hyperparameter selection=best-performing2026.02 | 53.78 | |
| DICEBackbone=VGG16, Method category=Confidence-based, Hyperparameter selection=best-performing2026.02 | 53.72 | |
| MCPBackbone=VGG16, Method category=Confidence-based, Hyperparameter selection=best-performing2026.02 | 53.59 | |
| WeiPerBackbone=VGG16, Method category=Confidence-based, Hyperparameter selection=best-performing2026.02 | 53.59 | |
| ReActBackbone=VGG16, Method category=Confidence-based, Hyperparameter selection=best-performing2026.02 | 51.79 | |
| DICEBackbone=ViT-B/32, Method Type=Confidence-based2026.02 | 51.54 | |
| FeatureNormBackbone=VGG16, Method category=Feature-based, Hyperparameter selection=best-performing2026.02 | 51.15 | |
| NANBackbone=VGG16, Method category=Feature-based, Hyperparameter selection=best-performing2026.02 | 50.26 | |
| NACBackbone=VGG16, Method category=Feature-based, Hyperparameter selection=best-performing2026.02 | 50.08 | |
| TAPUUDBackbone=ResNet18, Method Category=Feature-based, Hyperparameter Setting=best-performing2026.02 | 49.87 | |
| NuSABackbone=VGG16, Method category=Feature-based, Hyperparameter selection=best-performing2026.02 | 49.1 | |
| KDE (Gaussian)Backbone=ViT-B/32, Method Type=Feature-based2026.02 | 48.85 | |
| GAIA-ABackbone=ViT-B/32, Method Type=Confidence-based2026.02 | 44.29 | |
| NACBackbone=ViT-B/32, Method Type=Feature-based2026.02 | 43.21 | |
| GAIA-ABackbone=VGG16, Method category=Confidence-based, Hyperparameter selection=best-performing2026.02 | 41.73 | |
| ViMBackbone=ViT-B/32, Method Type=Confidence-based2026.02 | 37.44 | |
| ReActBackbone=ViT-B/32, Method Type=Confidence-based2026.02 | 34.42 | |
| GAIA-ABackbone=ResNet18, Method Category=Confidence-based, Hyperparameter Setting=best-performing2026.02 | 32.18 |