OOD Detection on CIFAR-10 OOD (test)
99.55AUROCHopfield Boosting
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Hopfield BoostingMethod type=OE, Augmentations=Weak, Auxiliary outlier data=true, Backbone=ResNet-182024.05 | 99.55 | 92 | |
| BMUncertainty Measure=Mutual Information (MI)2024.02 | 98.8 | — | |
| EDLUncertainty Measure=Differential Entropy (Dent)2024.02 | 98.8 | — | |
| EDLUncertainty Measure=Mutual Information (MI)2024.02 | 98.7 | — | |
| Bootstrap DistillUncertainty Measure=Differential Entropy (Dent)2024.02 | 98.4 | — | |
| BMUncertainty Measure=Differential Entropy (Dent)2024.02 | 98.3 | — | |
| Generalized (Neg)Backbone=ResNet18-SimCLR, ID Dataset=STL102026.03 | 98.3 | 5.5 | |
| Bootstrap DistillUncertainty Measure=Mutual Information (MI)2024.02 | 98.2 | — | |
| PostNetUncertainty Measure=Differential Entropy (Dent)2024.02 | 97.8 | — | |
| Fisher-EDLUncertainty Measure=Differential Entropy (Dent)2024.02 | 97.1 | — | |
| PALMMethod type=Training, Augmentations=Strong, Auxiliary outlier data=false, Backbone=ResNet-182024.05 | 97.02 | 1,591 | |
| END2Uncertainty Measure=Differential Entropy (Dent)2024.02 | 96.9 | — | |
| PostNetUncertainty Measure=Mutual Information (MI)2024.02 | 96.8 | — | |
| ASH-SBackbone=DenseNet-1012022.09 | 96.61 | 15.05 | |
| SSD+Method type=Training, Augmentations=Strong, Auxiliary outlier data=false, Backbone=ResNet-182024.05 | 96.57 | 2,035 | |
| NPOSMethod type=Training, Augmentations=Strong, Auxiliary outlier data=false, Backbone=ResNet-182024.05 | 96.31 | 2,144 | |
| ASH-BBackbone=DenseNet-1012022.09 | 96.02 | 20.23 | |
| END2Uncertainty Measure=Mutual Information (MI)2024.02 | 95.7 | — | |
| DICEBackbone=DenseNet-1012022.09 | 95.24 | 20.83 | |
| ASH-PBackbone=DenseNet-1012022.09 | 95.22 | 23.45 | |
| Fisher-EDLUncertainty Measure=Mutual Information (MI)2024.02 | 95.1 | — | |
| ReActBackbone=DenseNet-1012022.09 | 94.95 | 26.45 | |
| Energy scoreBackbone=DenseNet-1012022.09 | 94.57 | 26.55 | |
| S2DUncertainty Measure=Differential Entropy (Dent)2024.02 | 93.9 | — | |
| ODINBackbone=DenseNet-1012022.09 | 93.71 | 24.57 | |
| Generalized (Neg, InfoNCE)Backbone=ResNet18-SimCLR, ID Dataset=STL102026.03 | 92.8 | 39.8 | |
| GENMethod type=Post-hoc, Augmentations=Weak, Auxiliary outlier data=false, Backbone=ResNet-182024.05 | 92.65 | 3,507 | |
| EBOMethod type=Post-hoc, Augmentations=Weak, Auxiliary outlier data=false, Backbone=ResNet-182024.05 | 92.55 | 3,409 | |
| Softmax scoreBackbone=DenseNet-1012022.09 | 92.46 | 48.73 | |
| MaxLogitMethod type=Post-hoc, Augmentations=Weak, Auxiliary outlier data=false, Backbone=ResNet-182024.05 | 92.43 | 3,500 | |
| Single (Neg)Backbone=ResNet18-SimCLR, ID Dataset=STL102026.03 | 92 | 61 | |
| MSPMethod type=Post-hoc, Augmentations=Weak, Auxiliary outlier data=false, Backbone=ResNet-182024.05 | 91.64 | 4,921 | |
| ASHMethod type=Post-hoc, Augmentations=Weak, Auxiliary outlier data=false, Backbone=ResNet-182024.05 | 90.94 | 3,550 | |
| S2DUncertainty Measure=Mutual Information (MI)2024.02 | 90.3 | — | |
| MahalanobisBackbone=DenseNet-1012022.09 | 89.15 | 31.42 | |
| COMBOODBackbone=ResNet18-SimCLR, ID Dataset=STL102026.03 | 83.9 | 50 | |
| CIDERBackbone=ResNet18-SimCLR, ID Dataset=STL102026.03 | 83.9 | 43.9 | |
| RPriorNetUncertainty Measure=Differential Entropy (Dent)2024.02 | 81.3 | — | |
| RPriorNetUncertainty Measure=Mutual Information (MI)2024.02 | 77.6 | — | |
| D-KNNBackbone=ResNet18-SimCLR, ID Dataset=STL102026.03 | 76.9 | 56.2 | |
| MSPBackbone=ResNet18-SimCLR, ID Dataset=STL102026.03 | 71.8 | 66.1 | |
| Energy (T=1)Backbone=ResNet18-SimCLR, ID Dataset=STL102026.03 | 71 | 65.2 | |
| NatPNUncertainty Measure=Differential Entropy (Dent)2024.02 | 69.7 | — | |
| KNN (ℓ2)Backbone=ResNet18-SimCLR, ID Dataset=STL102026.03 | 57.7 | 85 | |
| MahalanobisBackbone=ResNet18-SimCLR, ID Dataset=STL102026.03 | 48.7 | 89.5 | |
| NatPNUncertainty Measure=Mutual Information (MI)2024.02 | 14.9 | — |