Image Classification on CIFAR-100 (test) (Uniform/Targeted Corruption Metrics)
63.65Accuracy (Clean)DR-IS
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| DR-ISBackbone=ResNet-18, Keep fraction (alpha)=0.252026.05 | 63.65 | 62.02 | 63.15 | 60.96 | |
| RandomBackbone=ResNet-18, Keep fraction (alpha)=0.252026.05 | 63.09 | 54.81 | 58.76 | 51.53 | |
| AUM (top-α)Backbone=ResNet-18, Keep fraction (alpha)=0.252026.05 | 55.32 | 55.14 | 54.5 | 53.32 | |
| EL2N (ensemble)Backbone=ResNet-18, Keep fraction (alpha)=0.252026.05 | 29.51 | 7.43 | 12.25 | 1.54 | |
| Consensus-lossBackbone=ResNet-18, Keep fraction (alpha)=0.252026.05 | 27.92 | 6.19 | 10.59 | 1.11 |