Out-of-Distribution Detection on CIFAR-100 (ID) vs CIFAR-10 (OOD)
79.78AUROCESOOD
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| ESOODBackbone=ResNet-182026.03 | 79.78 | — | — | 80.65 | |
| GENBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 79.38 | 58.87 | — | — | |
| WeiPer+MSPBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 79.24 | 59.69 | — | — | |
| MLSBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 79.21 | 59.11 | — | — | |
| WeiPer+KLDBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 79.2 | 59.9 | — | — | |
| EBOBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 79.05 | 59.21 | — | — | |
| TempScaleBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 79.02 | 58.72 | — | — | |
| WeiPer+ReActBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 79 | 60.41 | — | — | |
| ReActBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 78.65 | 61.3 | — | — | |
| MSPBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 78.47 | 58.91 | — | — | |
| ODINBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 78.18 | 60.64 | — | — | |
| ODINBackbone=ResNet182026.02 | 78.18 | — | 79.12 | — | |
| SHEBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 78.15 | 60.41 | — | — | |
| DICEBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 78.04 | 60.98 | — | — | |
| RMDSBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 77.75 | 61.37 | — | — | |
| NECOBackbone=DenseNet-1002026.03 | 77.12 | — | — | 63.5 | |
| MaxLogitBackbone=DenseNet-1002026.03 | 77.1 | — | — | 63.57 | |
| KNNBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 77.02 | 72.8 | — | — | |
| KNNBackbone=ResNet182026.02 | 77.02 | — | 75.36 | — | |
| EnergyBackbone=DenseNet-1002026.03 | 76.99 | — | — | 63.54 | |
| D_alpha (PBB)2021.06 | 76.8 | 64.2 | — | — | |
| ASHBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 76.48 | 68.06 | — | — | |
| NNGuideBackbone=ResNet-182026.03 | 75.82 | — | — | 79.33 | |
| MSPBackbone=DenseNet-1002026.03 | 75.72 | — | — | 62.15 | |
| MSPBackbone=ResNet-182026.03 | 75.62 | — | — | 77.26 | |
| ReActBackbone=DenseNet-1002026.03 | 75.46 | — | — | 68.24 | |
| COMBOODBackbone=ResNet182026.02 | 75.39 | — | 71.65 | — | |
| LaRExBackbone=DenseNet-1002026.03 | 75.12 | — | — | 70.05 | |
| Prototype FusionBackbone=ResNet-182026.03 | 74.47 | — | — | 79.84 | |
| OpenMaxBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 74.38 | 60.17 | — | — | |
| KLMBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 73.91 | 84.77 | — | — | |
| ESOODBackbone=DenseNet-1002026.03 | 73.45 | — | — | 65.67 | |
| ENERGYparameter setting=[23]2021.06 | 73.3 | 76.4 | — | — | |
| LaRExBackbone=ResNet-182026.03 | 73.11 | — | — | 82.45 | |
| NNGuideBackbone=DenseNet-1002026.03 | 72.79 | — | — | 83.51 | |
| MaxLogitBackbone=ResNet-182026.03 | 72.75 | — | — | 84.39 | |
| VIMBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 72.21 | 70.59 | — | — | |
| ViMBackbone=DenseNet-1002026.03 | 71.68 | — | — | 69.7 | |
| ODIN (OOD)parameter setting=[23]2021.06 | 70.5 | 79.5 | — | — | |
| MahalanobisBackbone=ResNet-182026.03 | 70.44 | — | — | 78.46 | |
| GradNormBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 70.32 | 84.3 | — | — | |
| Reg. MahalanobisBackbone=ResNet182026.02 | 68 | — | 63.19 | — | |
| Prototype FusionBackbone=DenseNet-1002026.03 | 67.57 | — | — | 79.15 | |
| EnergyBackbone=ResNet-182026.03 | 66.38 | — | — | 86.81 | |
| ViMBackbone=ResNet-182026.03 | 64.48 | — | — | 86.93 | |
| OpenGANBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 63.23 | 78.83 | — | — | |
| GradNormBackbone=DenseNet-1002026.03 | 60.23 | — | — | 90.05 | |
| RankFeatBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 58.04 | 82.78 | — | — | |
| MDSBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 55.87 | 88 | — | — | |
| MDSBackbone=ResNet182026.02 | 54.33 | — | 53.5 | — | |
| GramBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 49.41 | 92.71 | — | — | |
| GramBackbone=ResNet182026.02 | 46.99 | — | 46.02 | — | |
| MDSEnsBackbone=ResNet18, Training=Cross Entropy, Preprocessing=standard2024.05 | 43.85 | 95.94 | — | — | |
| MahalanobisBackbone=DenseNet-1002026.03 | 40.51 | — | — | 95.53 | |
| NECOBackbone=ResNet-182026.03 | 39.79 | — | — | 95.27 | |
| ReActBackbone=ResNet-182026.03 | 28.57 | — | — | 95.95 | |
| GradNormBackbone=ResNet-182026.03 | 27.77 | — | — | 97.37 |