Out-of-Distribution Detection on OBJECTS vs CIFAR-100-C (test)
82.66AUROCAENIB
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| AENIBScore=log Dir_0.05(y) + log N(zn; 0, I)2023.03 | 82.66 | 93.62 | 74.14 | |
| AENIBScore=log Dir_0.05(y)2023.03 | 79.39 | 92.33 | 81.51 | |
| SEMBase model=Cross-entropy2023.03 | 78.89 | 92.07 | 83.92 | |
| Cross-entropyScore=log Dir_0.05(y)2023.03 | 78.8 | 92.21 | 82.5 | |
| VIBScore=log Dir_0.05(y)2023.03 | 78.54 | 91.85 | 81.47 | |
| VIBScore=max p(y|x)2023.03 | 76.12 | 91.03 | 84.99 | |
| AENIBScore=max p(y|x)2023.03 | 74.31 | 89.89 | 86.26 | |
| Cross-entropyScore=max p(y|x)2023.03 | 74.12 | 89.74 | 87.26 | |
| Energy-basedBase model=Cross-entropy2023.03 | 68.37 | 85.54 | 83.64 | |
| MahalanobisBase model=Cross-entropy2023.03 | 68.13 | 82.97 | 85.53 | |
| ODINBase model=Cross-entropy2023.03 | 67.51 | 83.97 | 84.26 |