Out-of-Distribution Detection on OBJECTS vs Texture (test)
88.91AUROCAENIB
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| AENIBScore=log Dir_0.05(y) + log N(zn; 0, I)2023.03 | 88.91 | 97.49 | 48.44 | |
| SEMBase model=Cross-entropy2023.03 | 79.69 | 95.48 | 82.15 | |
| AENIBScore=log Dir_0.05(y)2023.03 | 75.04 | 93.83 | 86.01 | |
| AENIBScore=max p(y|x)2023.03 | 74.95 | 93.97 | 86.01 | |
| VIBScore=max p(y|x)2023.03 | 74.67 | 94.09 | 87.22 | |
| Cross-entropyScore=max p(y|x)2023.03 | 74.18 | 93.34 | 85.64 | |
| VIBScore=log Dir_0.05(y)2023.03 | 73.84 | 93.5 | 88 | |
| MahalanobisBase model=Cross-entropy2023.03 | 72.02 | 88.46 | 72.98 | |
| Cross-entropyScore=log Dir_0.05(y)2023.03 | 70.69 | 92.68 | 91.35 | |
| ODINBase model=Cross-entropy2023.03 | 70.14 | 89.97 | 72.91 | |
| Energy-basedBase model=Cross-entropy2023.03 | 68.63 | 89.51 | 75.57 |