Image Classification on ImageNette (test) (Adversarial Robustness Metrics)
86.9Clean AccuracyPAT-VR
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| PAT-VRBackbone=ResNet-50, Source Radius (epsilon)=0.5, lambda=0.05, Training=AlexNet-based PAT2022.04 | 86.9 | 34.9 | 40.6 | 9.4 | 64.6 | 3.7 | 21.9 | 2.2 | |
| PATBackbone=ResNet-50, Source Radius (epsilon)=0.5, lambda=0, Training=AlexNet-based PAT2022.04 | 86.6 | 38.8 | 44.3 | 5.8 | 60.8 | 2.1 | 16.2 | 2.2 | |
| PAT-VRBackbone=ResNet-50, Source Radius (epsilon)=0.5, lambda=0.1, Training=AlexNet-based PAT2022.04 | 85.1 | 31.4 | 37.1 | 44.9 | 80.5 | 24.9 | 48.7 | 29.7 | |
| PAT-VRBackbone=ResNet-50, Source Radius (epsilon)=1, lambda=0.1, Training=AlexNet-based PAT2022.04 | 72.5 | 29.4 | 35.1 | 61.8 | 70.7 | 28.8 | 56.9 | 30.8 | |
| PAT-VRBackbone=ResNet-50, Source Radius (epsilon)=1, lambda=0.05, Training=AlexNet-based PAT2022.04 | 72.1 | 29.5 | 34.8 | 59.6 | 69.7 | 28.2 | 56.7 | 18.5 | |
| PATBackbone=ResNet-50, Source Radius (epsilon)=1, lambda=0, Training=AlexNet-based PAT2022.04 | 71.6 | 28.7 | 33.3 | 64.5 | 67.5 | 27.8 | 26.6 | 9.8 |