Adversarial Attack on ImageNet (val) (Inception Model Robustness)
1Robust Accuracy (Inc-v3)NI-CT-FGSM
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| NI-CT-FGSMattack_type=gradient-based iterative, ensemble_generation=true, enhanced_by_CTM=true2021.03 | 1 | 1 | 1 | 1 | 0.928 | 0.896 | 0.836 | |
| VNI-CT-FGSMattack_type=gradient-based iterative, ensemble_generation=true, enhanced_by_CTM=true2021.03 | 1 | 0.999 | 0.996 | 1 | 0.955 | 0.945 | 0.923 | |
| MI-FGSMattack_type=gradient-based iterative, ensemble_generation=true2021.03 | 0.999 | 0.982 | 0.953 | 0.999 | 0.394 | 0.353 | 0.242 | |
| VNI-FGSMattack_type=gradient-based iterative, ensemble_generation=true2021.03 | 0.999 | 0.996 | 0.986 | 0.999 | 0.713 | 0.66 | 0.529 | |
| NI-FGSMattack_type=gradient-based iterative, ensemble_generation=true2021.03 | 0.998 | 0.998 | 0.989 | 0.998 | 0.41 | 0.335 | 0.231 | |
| VMI-FGSMattack_type=gradient-based iterative, ensemble_generation=true2021.03 | 0.997 | 0.985 | 0.96 | 0.999 | 0.676 | 0.629 | 0.507 | |
| VMI-CT-FGSMattack_type=gradient-based iterative, ensemble_generation=true, enhanced_by_CTM=true2021.03 | 0.997 | 0.992 | 0.984 | 0.999 | 0.936 | 0.924 | 0.91 | |
| MI-CT-FGSMattack_type=gradient-based iterative, ensemble_generation=true, enhanced_by_CTM=true2021.03 | 0.996 | 0.991 | 0.974 | 0.997 | 0.913 | 0.896 | 0.868 |