Adversarial Attack on ImageNet Adversarially Trained Models
76Robust Accuracy (Inc-v3 adv)NAA-PD
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| NAA-PDSource Model=Res-v22022.03 | 76 | 78.9 | 72.4 | 68 | 52.8 | |
| FIA-PDSource Model=Res-v22022.03 | 71 | 69.7 | 58.3 | 53.9 | 34.6 | |
| NAA-PDSource Model=IncRes-v22022.03 | 69.4 | 71.3 | 61.9 | 59 | 48.3 | |
| NAA-PDSource Model=Inc-v32022.03 | 67.9 | 68.6 | 55.4 | 55.6 | 33.8 | |
| FIA-PDSource Model=IncRes-v22022.03 | 66.9 | 66.7 | 49.7 | 44.9 | 31.9 | |
| NAA-PDSource Model=Inc-v42022.03 | 61.3 | 87.9 | 55.4 | 53.6 | 34.4 | |
| FIA-PDSource Model=Inc-v32022.03 | 59.8 | 57.1 | 38.5 | 37.3 | 21.5 | |
| FIA-PDSource Model=Inc-v42022.03 | 55.3 | 60.7 | 45.5 | 42.1 | 23.5 | |
| MIM-PDSource Model=Res-v22022.03 | 42 | 44.5 | 36.3 | 34.3 | 18.1 | |
| FDA-PDSource Model=IncRes-v22022.03 | 41.2 | 35.6 | 17.4 | 29.9 | 25.3 | |
| NRDM-PDSource Model=IncRes-v22022.03 | 40.8 | 32.3 | 18.6 | 30.6 | 26 | |
| MIM-PDSource Model=IncRes-v22022.03 | 36.7 | 41.7 | 28.8 | 26.7 | 16.3 | |
| FDA-PDSource Model=Inc-v42022.03 | 36.6 | 38 | 21.9 | 20.9 | 9.1 | |
| NRDM-PDSource Model=Inc-v42022.03 | 34.2 | 35 | 21.3 | 19.2 | 8.6 | |
| MIM-PDSource Model=Inc-v32022.03 | 31 | 28 | 21.3 | 22 | 9.3 | |
| MIM-PDSource Model=Inc-v42022.03 | 30.6 | 28.8 | 23.9 | 24.5 | 12.5 | |
| NRDM-PDSource Model=Inc-v32022.03 | 29.7 | 22.9 | 12.2 | 18.6 | 13.6 | |
| FDA-PDSource Model=Res-v22022.03 | 28.7 | 21.5 | 13.6 | 15.5 | 7.1 | |
| NRDM-PDSource Model=Res-v22022.03 | 26.2 | 18.2 | 13.8 | 14.5 | 5.9 | |
| FDA-PDSource Model=Inc-v32022.03 | 23 | 16 | 10.8 | 12.1 | 8 |