Out-of-distribution Detection on CIFAR-100 vs iSUN (test)
0.959TNR @ TPR95Ours
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| OursBackbone=DenseNet2019.12 | 0.959 | 0.99 | 0.956 | |
| OursArchitecture=ResNet, In-distribution Dataset=CIFAR-1002019.12 | 0.9512 | 0.989 | 0.9518 | |
| OursBackbone=ResNet2019.12 | 0.948 | 0.988 | 0.956 | |
| MahalanobisArchitecture=ResNet, In-distribution Dataset=CIFAR-1002019.12 | 0.8991 | 0.9791 | 0.9305 | |
| MahalanobisBackbone=ResNet2019.12 | 0.899 | 0.979 | 0.931 | |
| MahalanobisBackbone=DenseNet2019.12 | 0.87 | 0.974 | 0.924 | |
| VDArchitecture=ResNet, In-distribution Dataset=CIFAR-1002019.12 | 0.802 | 0.942 | 0.878 | |
| ODINArchitecture=ResNet, In-distribution Dataset=CIFAR-1002019.12 | 0.4521 | 0.8548 | 0.7847 | |
| ODINBackbone=ResNet2019.12 | 0.452 | 0.855 | 0.785 | |
| SemanticArchitecture=ResNet, In-distribution Dataset=CIFAR-1002019.12 | 0.416 | 0.852 | 0.884 | |
| ODINBackbone=DenseNet2019.12 | 0.374 | 0.845 | 0.764 | |
| BaselineBackbone=ResNet2019.12 | 0.169 | 0.758 | 0.701 | |
| BaselineArchitecture=ResNet, In-distribution Dataset=CIFAR-1002019.12 | 0.1689 | 0.758 | 0.7011 | |
| BaselineBackbone=DenseNet2019.12 | 0.149 | 0.695 | 0.638 |