OOD Detection on OpenImage-O
20.35FPR@95Maha++
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Maha++Backbone=EVA02-S14, ID Dataset=ImageNet-LT2026.06 | 20.35 | 96.08 | |
| rMaha++Backbone=EVA02-S14, ID Dataset=ImageNet-LT2026.06 | 20.35 | 96.08 | |
| MahaBackbone=EVA02-S14, ID Dataset=ImageNet-LT2026.06 | 21.36 | 95.92 | |
| rMahaBackbone=EVA02-S14, ID Dataset=ImageNet-LT2026.06 | 22.96 | 95.54 | |
| X-MahaBackbone=EVA02-S14, ID Dataset=ImageNet-LT2026.06 | 24.17 | 95.5 | |
| EPDBackbone=ViT-B/16, ID Accuracy=81.14%, ID Dataset=ImageNet-1K2026.04 | 26.03 | 92.12 | |
| OCOModel=DINOv22026.05 | 28.55 | 93.53 | |
| CoRPModel=DINOv22026.05 | 28.86 | 93.26 | |
| RMDSBackbone=ViT-B/16, ID Accuracy=81.14%, ID Dataset=ImageNet-1K2026.04 | 29.57 | 92.32 | |
| ViMBackbone=ViT-B/16, ID Accuracy=81.14%, ID Dataset=ImageNet-1K2026.04 | 29.59 | 92.18 | |
| FDBDModel=DINOv22026.05 | 30.3 | 91.89 | |
| MDSBackbone=ViT-B/16, ID Accuracy=81.14%, ID Dataset=ImageNet-1K2026.04 | 30.35 | 92.38 | |
| OODDModel=DINOv22026.05 | 31.25 | 91.54 | |
| OCOModel=ViT2026.05 | 32.76 | 91.22 | |
| SHEBackbone=ViT-B/16, ID Accuracy=81.14%, ID Dataset=ImageNet-1K2026.04 | 33.59 | 91.04 | |
| KNNBackbone=ViT-B/16, ID Accuracy=81.14%, ID Dataset=ImageNet-1K2026.04 | 34.82 | 89.86 | |
| VRA++Backbone=BiT, Pre-trained=ImageNet2023.02 | 34.94 | 93.55 | |
| GENBackbone=ViT-B/16, ID Accuracy=81.14%, ID Dataset=ImageNet-1K2026.04 | 35.43 | 90.27 | |
| FDBDModel=ViT2026.05 | 36.17 | 90.73 | |
| Maha++Backbone=Swin-T, In-Distribution (ID) Dataset=ImageNet-LT2026.06 | 36.86 | 93.14 | |
| rMaha++Backbone=Swin-T, In-Distribution (ID) Dataset=ImageNet-LT2026.06 | 37.03 | 93.15 | |
| NECOModel=DINOv22026.05 | 37.06 | 89.66 | |
| SCALEModel=DINOv22026.05 | 37.87 | 88.86 | |
| MaxLogitModel=DINOv22026.05 | 38.74 | 88.96 | |
| ReActBackbone=EVA02-S14, ID Dataset=ImageNet-LT2026.06 | 39.01 | 86.75 | |
| EnergyModel=DINOv22026.05 | 39.53 | 88.39 | |
| EnergyBackbone=EVA02-S14, ID Dataset=ImageNet-LT2026.06 | 39.72 | 86.28 | |
| MSPBackbone=EVA02-S14, ID Dataset=ImageNet-LT2026.06 | 41.19 | 89.49 | |
| MM++Backbone=Swin-T, In-Distribution (ID) Dataset=ImageNet-LT2026.06 | 43.12 | 92.44 | |
| rMahaBackbone=Swin-T, In-Distribution (ID) Dataset=ImageNet-LT2026.06 | 43.74 | 91.87 | |
| SCALEModel=ViT2026.05 | 43.81 | 87.9 | |
| ViMBackbone=BiT, Pre-trained=ImageNet2023.02 | 43.96 | 91.54 | |
| CoRPModel=ViT2026.05 | 44.55 | 86.71 | |
| MaxLogitModel=ViT2026.05 | 46.92 | 87.42 | |
| EnergyModel=ViT2026.05 | 47.49 | 86.89 | |
| SHEModel=DINOv22026.05 | 49.84 | 86.01 | |
| NNguideModel=DINOv22026.05 | 51.48 | 83.46 | |
| OODDModel=ViT2026.05 | 52.35 | 90.54 | |
| ODINBackbone=EVA02-S14, ID Dataset=ImageNet-LT2026.06 | 52.54 | 82.24 | |
| DICEBackbone=ViT-B/16, ID Accuracy=81.14%, ID Dataset=ImageNet-1K2026.04 | 52.57 | 82.23 | |
| KNNBackbone=EVA02-S14, ID Dataset=ImageNet-LT2026.06 | 53.32 | 87.05 | |
| MahaBackbone=Swin-T, In-Distribution (ID) Dataset=ImageNet-LT2026.06 | 54.3 | 90.47 | |
| ReActBackbone=BiT, Pre-trained=ImageNet2023.02 | 54.97 | 88.94 | |
| MSPBackbone=ViT-B/16, ID Accuracy=81.14%, ID Dataset=ImageNet-1K2026.04 | 56.11 | 84.87 | |
| NNguideModel=ViT2026.05 | 56.43 | 82.04 | |
| ReActBackbone=ViT-B/16, ID Accuracy=81.14%, ID Dataset=ImageNet-1K2026.04 | 57.68 | 84.29 | |
| TempScaleBackbone=ViT-B/16, ID Accuracy=81.14%, ID Dataset=ImageNet-1K2026.04 | 60 | 85.04 | |
| MSPBackbone=Swin-T, In-Distribution (ID) Dataset=ImageNet-LT2026.06 | 60.21 | 85.68 | |
| ReActBackbone=Swin-T, In-Distribution (ID) Dataset=ImageNet-LT2026.06 | 64.08 | 82.93 | |
| MahalanobisBackbone=BiT, Pre-trained=ImageNet2023.02 | 64.32 | 83.1 | |
| MM++Backbone=EVA02-S14, ID Dataset=ImageNet-LT2026.06 | 66.96 | 87.49 | |
| NECOModel=ViT2026.05 | 67.24 | 86.44 | |
| SHEModel=ViT2026.05 | 67.34 | 81.42 | |
| EnergyBackbone=Swin-T, In-Distribution (ID) Dataset=ImageNet-LT2026.06 | 67.98 | 79.08 | |
| ODINBackbone=BiT, Pre-trained=ImageNet2023.02 | 72.83 | 85.64 | |
| EnergyBackbone=BiT, Pre-trained=ImageNet2023.02 | 73.42 | 84.77 | |
| MSPBackbone=BiT, Pre-trained=ImageNet2023.02 | 73.72 | 84.16 | |
| OpenMaxBackbone=ViT-B/16, ID Accuracy=81.14%, ID Dataset=ImageNet-1K2026.04 | 73.74 | 87.36 | |
| ODINBackbone=Swin-T, In-Distribution (ID) Dataset=ImageNet-LT2026.06 | 81.28 | 72.12 | |
| MLSBackbone=ViT-B/16, ID Accuracy=81.14%, ID Dataset=ImageNet-1K2026.04 | 85.78 | 81.6 | |
| EBOBackbone=ViT-B/16, ID Accuracy=81.14%, ID Dataset=ImageNet-1K2026.04 | 88.79 | 76.48 | |
| KNNBackbone=Swin-T, In-Distribution (ID) Dataset=ImageNet-LT2026.06 | 89.41 | 54.11 | |
| GradNormBackbone=ViT-B/16, ID Accuracy=81.14%, ID Dataset=ImageNet-1K2026.04 | 94.52 | 37.82 | |
| ASHBackbone=ViT-B/16, ID Accuracy=81.14%, ID Dataset=ImageNet-1K2026.04 | 94.8 | 55.52 |