OOD detection on Foggy Cityscapes i=0.2
1AUROCOurs (SwinB-NF)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Ours (SwinB-NF)Category=VFMs, Backbone=SwinB, Modeling=NF2025.01 | 1 | 0 | |
| GradNormCategory=OOD Methods, Backbone=ResNet-502025.01 | 0.98 | 0.06 | |
| RN50-NFCategory=Image-Net Trained Counterparts, Backbone=ResNet-50, Modeling=NF2025.01 | 0.93 | 0.25 | |
| RN50-GMMCategory=Image-Net Trained Counterparts, Backbone=ResNet-50, Modeling=GMM2025.01 | 0.92 | 0.28 | |
| SwinB-NFCategory=Image-Net Trained Counterparts, Backbone=SwinB, Modeling=NF2025.01 | 0.92 | 0.35 | |
| Ours (RN50-NF)Category=VFMs, Backbone=RN50, Modeling=NF2025.01 | 0.86 | 0.53 | |
| Ours (RN50-GMM)Category=VFMs, Backbone=RN50, Modeling=GMM2025.01 | 0.85 | 0.47 | |
| DICECategory=OOD Methods, Backbone=ResNet-502025.01 | 0.69 | 0.92 | |
| EntropyCategory=OOD Methods, Backbone=ResNet-502025.01 | 0.64 | 0.9 | |
| MaxLogitCategory=OOD Methods, Backbone=ResNet-502025.01 | 0.63 | 0.95 | |
| ODINCategory=OOD Methods, Backbone=ResNet-502025.01 | 0.59 | 0.93 | |
| KNNCategory=OOD Methods, Backbone=ResNet-502025.01 | 0.57 | 0.88 | |
| EnergyCategory=OOD Methods, Backbone=ResNet-502025.01 | 0.49 | 0.99 |