Learning-to-Defer on DermaMNIST (test)
83.39Classification AccuracyMao et al.
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Mao et al.Backbone=ResNet-18, Attack=PGD, epsilon=0.031372025.10 | 83.39 | 30.82 | 27.08 | 69.6 | |
| OursBackbone=ResNet-18, Attack=PGD, epsilon=0.03137, rho=1.75, nu=0.0012025.10 | 81.8 | 71.12 | 80.65 | 31.66 |