Out-of-Distribution Detection on ImageNet-C (AUROC, AUPR, FPR95)
99.36AUROCSupervised classifier
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Supervised classifierbackbone=ResNet502026.05 | 99.36 | 98.86 | 2.48 | |
| S-PUNAbackbone=ResNet50, classifier_usage=with classifier2026.05 | 99.34 | 99.89 | 2.59 | |
| S-PUNAbackbone=ResNet50, classifier_usage=without classifier2026.05 | 97.18 | 99.56 | 8.64 | |
| LaGAMbackbone=ResNet502026.05 | 95.07 | 89.85 | 24.01 | |
| saPUbackbone=ResNet502026.05 | 93.48 | 84.86 | 38.11 | |
| Dist-PUbackbone=ResNet502026.05 | 93.29 | 88.63 | 30.31 | |
| MM++Backbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 82.52 | — | 56.39 | |
| Maha++Backbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 79.27 | — | 64.26 | |
| rMaha++Backbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 79.24 | — | 64.32 | |
| MSPBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 78.65 | — | 60.8 | |
| KNNbackbone=ResNet502026.05 | 78.11 | 77.55 | 52.29 | |
| rMahaBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 77.74 | — | 69.75 | |
| Dc-PUbackbone=ResNet502026.05 | 75.74 | 58.96 | 72.3 | |
| MahaBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 75.59 | — | 74.85 | |
| ReActBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 73.33 | — | 77.46 | |
| EnergyBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 70.11 | — | 80.46 | |
| ODINBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 69.32 | — | 80.01 | |
| KNNBackbone=ConvNeXt-T, In-Distribution Dataset=ImageNet-LT2026.06 | 59.54 | — | 84.86 |