OOD Detection on Textures (OOD)
97.6AUROCASH-S
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ASH-SBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.09 | 97.6 | 11.93 | — | |
| ASH-SBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.09 | 97.1 | 13.12 | — | |
| DICE + ReActBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.09 | 96.25 | 16.28 | — | |
| ASH-BBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.09 | 95.5 | 21.17 | — | |
| ASH-BBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.09 | 95.07 | 20.92 | — | |
| LINePost-hoc=true2023.03 | 94.44 | 22.54 | — | |
| DICE + ReActPost-hoc=true2023.03 | 92.74 | 28.07 | — | |
| DICE + ReActBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.09 | 92.74 | 28.07 | — | |
| ReActBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.09 | 91.53 | 38.42 | — | |
| DICEBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.09 | 91.3 | 32.8 | — | |
| DICEPost-hoc=true2023.03 | 90.3 | 31.72 | — | |
| DICEBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.09 | 90.3 | 31.72 | — | |
| ReActPost-hoc=true2023.03 | 89.8 | 47.3 | — | |
| ReActBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.09 | 89.8 | 47.3 | — | |
| ASH-PBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.09 | 89.7 | 42.06 | — | |
| ASH-PBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.09 | 88.72 | 48.48 | — | |
| EnergyBackbone=ViT-B/16, In-Distribution Dataset=ImageNet-1k2025.05 | 87.8 | 47.87 | — | |
| ReActBackbone=ViT-B/16, In-Distribution Dataset=ImageNet-1k2025.05 | 87.65 | 45.69 | — | |
| KNNBackbone=ViT-B/16, In-Distribution Dataset=ImageNet-1k2025.05 | 87.57 | 50.3 | — | |
| KL Matching2021.05 | 87.07 | 49.7 | 97.97 | |
| DICEBackbone=ViT-B/16, In-Distribution Dataset=ImageNet-1k2025.05 | 86.96 | 60.2 | — | |
| Max logitsBackbone=ViT-B/16, In-Distribution Dataset=ImageNet-1k2025.05 | 86.87 | 52.5 | — | |
| ODINBackbone=ViT-B/16, In-Distribution Dataset=ImageNet-1k2025.05 | 86.87 | 52.49 | — | |
| MahalanobisBackbone=ViT-B/16, In-Distribution Dataset=ImageNet-1k2025.05 | 86.62 | 54.3 | — | |
| GradPCABackbone=ViT-B/16, In-Distribution Dataset=ImageNet-1k2025.05 | 86.21 | 59.94 | — | |
| EnergyPost-hoc=true2023.03 | 85.99 | 53.72 | — | |
| Energy scoreBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.09 | 85.99 | 53.72 | — | |
| ODINPost-hoc=true2023.03 | 85.62 | 50.23 | — | |
| ODINBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.09 | 85.62 | 50.23 | — | |
| ODINBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.09 | 85.03 | 49.96 | — | |
| Energy scoreBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.09 | 85.03 | 58.05 | — | |
| MahalanobisPost-hoc=true2023.03 | 85.01 | 55.8 | — | |
| MahalanobisBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.09 | 85.01 | 55.8 | — | |
| MOSgrouping=taxonomy-based2021.05 | 81.23 | 60.43 | 96.65 | |
| MSPBackbone=ViT-B/16, In-Distribution Dataset=ImageNet-1k2025.05 | 80 | 62.07 | — | |
| MSPPost-hoc=true2023.03 | 79.61 | 68 | — | |
| Softmax scoreBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.09 | 79.61 | 68 | — | |
| Softmax scoreBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.09 | 77.3 | 73.51 | — | |
| ODIN2021.05 | 76.3 | 81.31 | 96.12 | |
| Energy2021.05 | 75.79 | 80.87 | 96.05 | |
| MSP2021.05 | 74.45 | 82.73 | 95.65 | |
| Mahalanobis2021.05 | 72.1 | 52.23 | 91.89 | |
| PS-CDTraining Set=CIFAR-102021.11 | 44 | — | — | |
| CDTraining Set=CIFAR-102021.11 | 36 | — | — | |
| MahalanobisBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.09 | 33.06 | 92.38 | — | |
| PixelCNN++Training Set=CIFAR-102021.11 | 33 | — | — | |
| GlowTraining Set=CIFAR-102021.11 | 27 | — | — |