Out-of-Distribution Detection on ImageNet-1k vs NINCO
89.9AUROCFDBD
Evaluation results
| Method | Links | ||
|---|---|---|---|
| FDBDBackbone=ResNet-502026.02 | 89.9 | 41.6 | |
| ReActBackbone=ResNet-502026.02 | 89.3 | 41.8 | |
| D-KNNBackbone=ResNet-502026.02 | 89.1 | 40.5 | |
| DINOv2+MLSBackbone=DINOv2, Training=Linear Probe, Postprocessor=MLS2024.10 | 88.38 | 41.02 | |
| ViMBackbone=ResNet-502026.02 | 87.8 | 48.9 | |
| EnergyBackbone=ResNet-502026.02 | 87.5 | 49.8 | |
| MaxLogitBackbone=ResNet-502026.02 | 87.5 | 51.5 | |
| ViT-B+CE+RMDSBackbone=ViT-B, Training=Cross Entropy (CE), Postprocessor=RMDS2024.10 | 87.31 | 46.2 | |
| KNNBackbone=ResNet-502026.02 | 86.1 | 42.5 | |
| GENBackbone=Swin-T, ID Accuracy=81.60%2026.04 | 85.16 | 48.1 | |
| RMDSBackbone=Swin-T, ID Accuracy=81.60%2026.04 | 84.91 | 49.89 | |
| MSPBackbone=ResNet-502026.02 | 83.8 | 60.9 | |
| Energy Profile Divergence (EPD)Backbone=Swin-T, ID Accuracy=81.60%2026.04 | 83.01 | 55.4 | |
| SHEBackbone=Swin-T, ID Accuracy=81.60%2026.04 | 82.74 | 67.53 | |
| ReActBackbone=Swin-T, ID Accuracy=81.60%2026.04 | 82.12 | 60.08 | |
| TempScaleBackbone=Swin-T, ID Accuracy=81.60%2026.04 | 82.09 | 63.02 | |
| ViMBackbone=Swin-T, ID Accuracy=81.60%2026.04 | 81.85 | 60.8 | |
| CoOD*Base Model=Standard2026.04 | 81.8 | 61.4 | |
| MDSBackbone=Swin-T, ID Accuracy=81.60%2026.04 | 81.78 | 53.55 | |
| MSPBackbone=Swin-T, ID Accuracy=81.60%2026.04 | 81.69 | 60.36 | |
| OpenMaxBackbone=Swin-T, ID Accuracy=81.60%2026.04 | 81.69 | 72.85 | |
| ID-Like + CoOD*Base Model=ID-Like2026.04 | 81.2 | 60 | |
| CoOp + CoOD*Base Model=CoOp2026.04 | 81.1 | 69.3 | |
| MLSBackbone=Swin-T, ID Accuracy=81.60%2026.04 | 80.95 | 75.78 | |
| CoOp + CoODBase Model=CoOp2026.04 | 80.6 | 52.2 | |
| CoODBase Model=Standard2026.04 | 80 | 65.8 | |
| ID-Like + CoODBase Model=ID-Like2026.04 | 79.4 | 63.9 | |
| KNNBackbone=Swin-T, ID Accuracy=81.60%2026.04 | 79.16 | 58.14 | |
| EBOBackbone=Swin-T, ID Accuracy=81.60%2026.04 | 78.42 | 79.23 | |
| GapBase Model=Standard2026.04 | 76.5 | 77.4 | |
| GLBase Model=Standard2026.04 | 76 | 74.4 | |
| ID-Like + GapBase Model=ID-Like2026.04 | 75.4 | 77.5 | |
| WDiscOODBackbone=ResNet-502026.02 | 74.7 | 68.5 | |
| CoOD-FBase Model=Standard2026.04 | 74.4 | 75.6 | |
| MLSBase Model=Standard2026.04 | 74.3 | 79.4 | |
| NecoBackbone=ResNet-502026.02 | 74 | 59.4 | |
| CoOp + GapBase Model=CoOp2026.04 | 74 | 80.7 | |
| MLS + TAGBase Model=Standard2026.04 | 73.9 | 77.7 | |
| MCMBase Model=Standard2026.04 | 73.6 | 79.6 | |
| EBOBase Model=Standard2026.04 | 72 | 84.1 | |
| ID-LikeBase Model=ID-Like2026.04 | 71.8 | 81.2 | |
| MahalanobisBackbone=ResNet-502026.02 | 71.7 | 70.9 | |
| MCM + TAGBase Model=Standard2026.04 | 71.3 | 81.6 | |
| ID-Like + CoOD-FBase Model=ID-Like2026.04 | 71.2 | 74.4 | |
| EBO + TAGBase Model=Standard2026.04 | 71.1 | 83.1 | |
| CoOpBase Model=CoOp2026.04 | 70.2 | 81.5 | |
| CoOp + CoOD-FBase Model=CoOp2026.04 | 70.2 | 81.5 | |
| DICEBackbone=Swin-T, ID Accuracy=81.60%2026.04 | 50 | 97.28 | |
| ASHBackbone=Swin-T, ID Accuracy=81.60%2026.04 | 47 | 93.98 | |
| GradNormBackbone=Swin-T, ID Accuracy=81.60%2026.04 | 45.01 | 94.06 |