OOD Detection on ImageNet-O
0.8421AUROCVRA++
Evaluation results
| Method | Links | ||
|---|---|---|---|
| VRA++Backbone=BiT, Pre-trained=ImageNet2023.02 | 0.8421 | 0.6045 | |
| ViMBackbone=BiT, Pre-trained=ImageNet2023.02 | 0.8387 | 0.615 | |
| CADetTraining=Self-supervised (contrastive), Backbone=ResNet502022.10 | 0.8229 | — | |
| ResidualTraining=Supervised, Backbone=ResNet502022.10 | 0.8115 | — | |
| ViMTraining=Supervised, Backbone=ResNet502022.10 | 0.8102 | — | |
| MahalanobisBackbone=BiT, Pre-trained=ImageNet2023.02 | 0.8037 | 0.7005 | |
| MahalanobisTraining=Supervised, Backbone=ResNet502022.10 | 0.8015 | — | |
| CADetTraining=Supervised, Backbone=ResNet502022.10 | 0.7073 | — | |
| ReActTraining=Supervised, Backbone=ResNet502022.10 | 0.6802 | — | |
| ReActBackbone=BiT, Pre-trained=ImageNet2023.02 | 0.6707 | 0.917 | |
| KL MatchingTraining=Supervised, Backbone=ResNet502022.10 | 0.67 | — | |
| EnergyBackbone=BiT, Pre-trained=ImageNet2023.02 | 0.6359 | 0.964 | |
| ODINBackbone=BiT, Pre-trained=ImageNet2023.02 | 0.63 | 0.9685 | |
| MSPBackbone=BiT, Pre-trained=ImageNet2023.02 | 0.5712 | 0.9685 | |
| MSPTraining=Supervised, Backbone=ResNet502022.10 | 0.5613 | — | |
| MaxLogitTraining=Supervised, Backbone=ResNet502022.10 | 0.5439 | — | |
| EnergyTraining=Supervised, Backbone=ResNet502022.10 | 0.5395 | — | |
| ODINTraining=Supervised, Backbone=ResNet502022.10 | 0.5287 | — |