Out-of-Distribution Detection on CelebA (ID) vs CIFAR-10 (OOD) (test)
0.9999AUROCSPEM
Evaluation Results
| Method | Links | |
|---|---|---|
| SPEMFlow Model=ResFlow, Feature Extractor=ResNet-152, Hyperparameter alpha=0.42026.02 | 0.9999 | |
| SPEM-noiseFlow Model=ResFlow, Feature Extractor=ResNet-152, Hyperparameter alpha=0.12026.02 | 0.9999 | |
| lambda (Similarity-only)2026.02 | 0.9999 | |
| LIDFlow Model=ResFlow2026.02 | 0.939 | |
| SITN2026.05 | 0.91 | |
| SITN2026.05 | 0.91 | |
| Log-likelihood2026.05 | 0.728 | |
| Log-likelihood2026.05 | 0.728 | |
| DoSE2026.05 | 0.717 | |
| Typicality2026.05 | 0.688 | |
| GMMFlow Model=ResFlow2026.02 | 0.6845 | |
| Typicality (sqrt(d))Flow Model=ResFlow2026.02 | 0.6628 | |
| WAICensemble size=52026.05 | 0.609 | |
| Typicality (Entropy)Flow Model=ResFlow2026.02 | 0.6082 | |
| LikelihoodFlow Model=ResFlow2026.02 | 0.5819 | |
| ComplexityFlow Model=ResFlow2026.02 | 0.574 | |
| Complexity2026.05 | 0.224 | |
| Likelihood RatioFlow Model=ResFlow2026.02 | 0.1826 |