Out-of-Distribution Detection on ImageNet V2
88.1AUROCSupervised classifier
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Supervised classifierbackbone=ResNet502026.05 | 88.1 | 92.07 | 51.56 | |
| S-PUNAbackbone=ResNet50, classifier_usage=with classifier2026.05 | 84.64 | 89.21 | 64.26 | |
| S-PUNAbackbone=ResNet50, classifier_usage=without classifier2026.05 | 70.22 | 80.54 | 80.64 | |
| MSPBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 61.53 | — | 88.49 | |
| ReActBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 60.06 | — | 89.37 | |
| ODINBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 59.71 | — | 90.68 | |
| MM++Backbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 59.7 | — | 90.33 | |
| rMahaBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 59.4 | — | 90.43 | |
| rMaha++Backbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 59.32 | — | 90.32 | |
| Maha++Backbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 59.29 | — | 90.32 | |
| EnergyBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 58.86 | — | 89.73 | |
| MahaBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 57.61 | — | 91.22 | |
| Dc-PUbackbone=ResNet502026.05 | 55.72 | 68.86 | 93.32 | |
| saPUbackbone=ResNet502026.05 | 54.3 | 68.75 | 92.32 | |
| LaGAMbackbone=ResNet502026.05 | 53.8 | 71.43 | 91.89 | |
| Dist-PUbackbone=ResNet502026.05 | 50.65 | 65.63 | 94.69 | |
| KNNbackbone=ResNet502026.05 | 50.28 | 65.75 | 94.43 | |
| KNNBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 46.21 | — | 96.34 |