Image Classification and OOD Detection on MNIST (ID) vs FashionMNIST (OOD) (test)
99.98ID Accuracy (%)C-EDL
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| C-EDLUncertainty Strategy=Monte Carlo2025.06 | 99.98 | 92.14 | 1.77 | 21.62 | |
| EDL++Uncertainty Strategy=Monte Carlo2025.06 | 99.97 | 93.35 | 7.55 | 38.81 | |
| EDL++Uncertainty Strategy=Metamorphic Transformation2025.06 | 99.97 | 94.93 | 1.96 | 16.43 | |
| Posterior Network2025.06 | 99.96 | 94.72 | 3.55 | 61.14 | |
| EDL2025.06 | 99.96 | 96.61 | 2.52 | 52.21 | |
| H-EDL2025.06 | 99.96 | 96.18 | 2.28 | 50.4 | |
| R-EDL2025.06 | 99.96 | 96.27 | 2.34 | 49.05 | |
| C-EDLUncertainty Strategy=Metamorphic Transformation2025.06 | 99.96 | 94.18 | 2 | 15.51 | |
| I-EDL2025.06 | 99.95 | 96.08 | 2.74 | 47.58 | |
| S-EDL2025.06 | 99.95 | 96.06 | 2.41 | 48.8 | |
| DA-EDL2025.06 | 99.89 | 95.16 | 2.98 | 28.74 |