Out-of-Distribution Detection on NINCO
87.31AUROCRMDS
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| RMDSBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 87.31 | 46.2 | — | |
| MM++Backbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 86.97 | 55.61 | — | |
| rMaha++Backbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 86.68 | 56.45 | — | |
| Maha++Backbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 86.62 | 56.49 | — | |
| MDSBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 86.52 | 48.77 | — | |
| rMahaBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 85.43 | 66.01 | — | |
| WeiPer+KLDBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 85.37 | 48.67 | — | |
| VIMBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 84.64 | 57.41 | — | |
| SHEBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 84.18 | 56.02 | — | |
| ASHBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 83.45 | 52.97 | — | |
| GENBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 82.51 | 59.33 | — | |
| WeiPer+ReActBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 82.49 | 53.36 | — | |
| WeiPer+MSPBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 82.35 | 52.53 | — | |
| KNNBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 82.25 | 54.73 | — | |
| RMDSBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 82.22 | 52.2 | — | |
| MahaBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 82.11 | 74.49 | — | |
| KLMBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 81.9 | 60.36 | — | |
| ReActBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 81.73 | 55.82 | — | |
| WeiPer+KLDBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 81.73 | 60.45 | — | |
| GENBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 81.7 | 54.9 | — | |
| TempScaleBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 81.41 | 55.1 | — | |
| WeiPer+ReActBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 81.07 | 62.67 | — | |
| KLMBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 80.68 | 66.14 | — | |
| MSPBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 80.68 | 65.45 | — | |
| WeiPer+MSPBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 80.66 | 64.85 | — | |
| MLSBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 80.41 | 59.44 | — | |
| MSPBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 79.95 | 56.88 | — | |
| EBOBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 79.7 | 60.58 | — | |
| KNNBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 79.64 | 58.39 | — | |
| OpenMaxBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 78.68 | 88.33 | — | |
| VIMBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 78.63 | 62.29 | — | |
| Reference Masking RegularizationModel=SWINv2-T2024.06 | 78.49 | — | — | |
| Reference Masking RegularizationModel=SWINV2-S2024.06 | 78.47 | — | — | |
| OpenMaxBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 78.17 | 60.81 | — | |
| MSPBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 78.11 | 77.28 | — | |
| TempScaleBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 77.8 | 81.88 | — | |
| ODINBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 77.77 | 68.16 | — | |
| Standard TrainingModel=SWINv2-T2024.06 | 77.46 | — | — | |
| SHEBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 76.49 | 69.72 | — | |
| DICEBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 76.01 | 66.9 | — | |
| ReActBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 75.43 | 78.51 | — | |
| Standard TrainingModel=SWINV2-S2024.06 | 74.73 | — | — | |
| Reference Masking RegularizationModel=SWINv2-B2024.06 | 74.53 | — | — | |
| GradNormBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 74.02 | 79.54 | — | |
| MLSBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 72.4 | 92.97 | — | |
| Standard TrainingModel=SWINv2-B2024.06 | 72.13 | — | — | |
| DICEBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 71.67 | 81.1 | — | |
| Reference Masking RegularizationModel=VITb162024.06 | 70.32 | — | — | |
| ReActBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 68.02 | 85.23 | — | |
| EBOBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 66.02 | 94.14 | — | |
| GramBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 66.01 | 83.87 | — | |
| Standard TrainingModel=VITb162024.06 | 65.98 | — | — | |
| EnergyBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 63.71 | 87.11 | — | |
| GramIn-Distribution Dataset=ImageNet-2002026.03 | 63.35 | 87.59 | 19.31 | |
| MDSBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 62.38 | 78.8 | — | |
| ODINBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 61.74 | 88.38 | — | |
| RMDSIn-Distribution Dataset=ImageNet-2002026.03 | 60.95 | 79.23 | 15.34 | |
| KLMIn-Distribution Dataset=ImageNet-2002026.03 | 60.52 | 85.33 | 15.72 | |
| ASHIn-Distribution Dataset=ImageNet-2002026.03 | 58.51 | 84.21 | 14.53 | |
| MSPIn-Distribution Dataset=ImageNet-2002026.03 | 57.76 | 79.3 | 13.24 | |
| VIMIn-Distribution Dataset=ImageNet-2002026.03 | 57.09 | 82.35 | 13.85 | |
| RAS HLIn-Distribution Dataset=ImageNet-2002026.03 | 56.9 | 80.56 | 13.31 | |
| TempScaleIn-Distribution Dataset=ImageNet-2002026.03 | 56.86 | 79.51 | 12.95 | |
| MDSIn-Distribution Dataset=ImageNet-2002026.03 | 56.65 | 88.52 | 14.22 | |
| SHEIn-Distribution Dataset=ImageNet-2002026.03 | 56.64 | 86.49 | 13.78 | |
| RASIn-Distribution Dataset=ImageNet-2002026.03 | 56.56 | 80.03 | 13.07 | |
| MDSEnsBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 55.41 | 91.86 | — | |
| GradNormIn-Distribution Dataset=ImageNet-2002026.03 | 55.32 | 91.68 | 13.43 | |
| GENIn-Distribution Dataset=ImageNet-2002026.03 | 54.85 | 80.12 | 12.37 | |
| KNNIn-Distribution Dataset=ImageNet-2002026.03 | 54.51 | 81.23 | 12.54 | |
| MLSIn-Distribution Dataset=ImageNet-2002026.03 | 54.43 | 83.27 | 12.34 | |
| DICEIn-Distribution Dataset=ImageNet-2002026.03 | 54.31 | 84.24 | 12.72 | |
| OpenMaxIn-Distribution Dataset=ImageNet-2002026.03 | 54.15 | 85 | 13.08 | |
| ReActIn-Distribution Dataset=ImageNet-2002026.03 | 53.84 | 84.61 | 12.33 | |
| EBOIn-Distribution Dataset=ImageNet-2002026.03 | 53.45 | 83.61 | 12.08 | |
| ASHBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 52.51 | 95.37 | — | |
| ODINIn-Distribution Dataset=ImageNet-2002026.03 | 52.36 | 86.66 | 11.63 | |
| OpenGANIn-Distribution Dataset=ImageNet-2002026.03 | 46.51 | 94.47 | 11.04 | |
| RankFeatBackbone=ResNet50, Training Loss=Cross Entropy, Preprocessing=Standard Preprocessing2024.05 | 46.08 | 94.03 | — | |
| RankFeatIn-Distribution Dataset=ImageNet-2002026.03 | 43.03 | 96.98 | 10.33 | |
| MDSEnsIn-Distribution Dataset=ImageNet-2002026.03 | 41.58 | 95.83 | 9.46 | |
| GradNormBackbone=ViT16-B, Training Objective=Cross Entropy, Preprocessing=standard2024.05 | 35.6 | 95.81 | — | |
| KNNBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 34.64 | 99.07 | — | |
| DisCoPatch-64Number of Patches=642025.01 | 0.943 | 0.39 | — | |
| NNGuideBackbone=RegNet2025.01 | 0.937 | 0.289 | — | |
| MOODv2Backbone=BEiTv22025.01 | 0.927 | 0.382 | — | |
| RankFeatBackbone=ResNetv2-1012025.01 | 0.9 | 0.393 | — | |
| SCALEBackbone=ResNet-502025.01 | 0.854 | 0.518 | — | |
| ASHBackbone=ResNet-502025.01 | 0.83 | 0.642 | — | |
| FDBDBackbone=ResNet-502025.01 | 0.814 | 0.661 | — | |
| NNGuideBackbone=ResNet-502025.01 | 0.805 | 0.698 | — | |
| NNGuideBackbone=MobileNetV22025.01 | 0.755 | 0.802 | — |