Out-of-distribution detection on CIFAR100 (0-49) ID vs Mixed OOD Set (test)
92.91AUROCOne-vs-All Oracle
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| One-vs-All Oracle2022.01 | 92.91 | 47.68 | 99.39 | 22.35 | |
| Entropy Oracle2022.01 | 86.16 | 38.63 | 98.68 | 44.67 | |
| UQGANMC-Dropout=true2022.01 | 80.75 | 31.75 | 98.04 | 58.1 | |
| UQGANMC-Dropout=false2022.01 | 80.11 | 28.81 | 97.98 | 55.23 | |
| GEN2022.01 | 77.43 | 26.12 | 97.64 | 61.48 | |
| Deep-Ensembles2022.01 | 74.29 | 29.38 | 96.53 | 83.37 | |
| Bayes-by-Backprop2022.01 | 69.74 | 24.6 | 95.86 | 87.01 | |
| Entropy2022.01 | 69.43 | 23.75 | 95.77 | 87.08 | |
| Confident Classifier2022.01 | 68.66 | 22.51 | 95.75 | 86.17 | |
| MC-Dropout2022.01 | 67.75 | 22.31 | 95.4 | 89.38 | |
| Max. Softmax2022.01 | 67.68 | 23.03 | 95.47 | 88.42 | |
| One-vs-All Baseline2022.01 | 62.99 | 15.69 | 94.53 | 92.44 |