Image Classification on SVHN (test) (Gradient Sensitivity Metrics)
96.37AccuracyAdversarial Training
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Adversarial TrainingBackbone=VGG-112022.06 | 96.37 | 1.25 | 27.64 | 24.23 | |
| LCNNs + GradRegBackbone=VGG-112022.06 | 96.23 | 2.02 | 25.54 | 17.06 | |
| GradRegBackbone=VGG-112022.06 | 96.03 | 1.85 | 57.52 | 33.34 | |
| StandardBackbone=VGG-112022.06 | 96.01 | 2.87 | 158.29 | 54.24 | |
| LCNNsBackbone=VGG-112022.06 | 95.61 | 4.04 | 83.34 | 30.05 |