Privacy-Preserving Image Classification on TinyImageNet (train)
0.0025Dec.-based MSENo_defense
Evaluation results
| Method | Links | ||
|---|---|---|---|
| No_defenseBackbone=ResNet-202025.03 | 0.0025 | 0.002 | |
| BottleneckBackbone=ResNet-202025.03 | 0.0107 | 0.0091 | |
| Bottleneck+CEMBackbone=ResNet-202025.03 | 0.014 | 0.0123 | |
| DistCorrBackbone=ResNet-202025.03 | 0.0148 | 0.0126 | |
| Noise_NopeekBackbone=ResNet-202025.03 | 0.0161 | 0.0147 | |
| DropoutBackbone=ResNet-202025.03 | 0.0165 | 0.0148 | |
| ResSFLBackbone=ResNet-202025.03 | 0.0173 | 0.0157 | |
| Noise_Nopeek+CEMBackbone=ResNet-202025.03 | 0.0183 | 0.0166 | |
| PATROLBackbone=ResNet-202025.03 | 0.0187 | 0.0168 | |
| Dropout+CEMBackbone=ResNet-202025.03 | 0.0188 | 0.0167 | |
| DistCorr+CEMBackbone=ResNet-202025.03 | 0.0195 | 0.0167 | |
| ResSFL+CEMBackbone=ResNet-202025.03 | 0.0197 | 0.018 | |
| PATROL+CEMBackbone=ResNet-202025.03 | 0.0211 | 0.0185 | |
| Noise_ARLBackbone=ResNet-202025.03 | 0.0229 | 0.0204 | |
| Noise_ARL+CEMBackbone=ResNet-202025.03 | 0.0281 | 0.0251 |