Privacy-Preserving Image Classification on TinyImageNet (inference)
53.73AccuracyNo_defense
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| No_defenseBackbone=ResNet-202025.03 | 53.73 | 0.0022 | 0.0021 | |
| BottleneckBackbone=ResNet-202025.03 | 52.77 | 0.0103 | 0.0092 | |
| Bottleneck+CEMBackbone=ResNet-202025.03 | 52.25 | 0.0136 | 0.0125 | |
| ResSFL+CEMBackbone=ResNet-202025.03 | 52.07 | 0.0194 | 0.0173 | |
| Noise_Nopeek+CEMBackbone=ResNet-202025.03 | 52.01 | 0.018 | 0.0167 | |
| ResSFLBackbone=ResNet-202025.03 | 51.99 | 0.0172 | 0.0161 | |
| DistCorrBackbone=ResNet-202025.03 | 51.79 | 0.0145 | 0.0136 | |
| DistCorr+CEMBackbone=ResNet-202025.03 | 51.78 | 0.019 | 0.0168 | |
| PATROLBackbone=ResNet-202025.03 | 51.75 | 0.0187 | 0.0176 | |
| PATROL+CEMBackbone=ResNet-202025.03 | 51.64 | 0.0209 | 0.0189 | |
| Noise_NopeekBackbone=ResNet-202025.03 | 51.63 | 0.0157 | 0.0152 | |
| Noise_ARLBackbone=ResNet-202025.03 | 51.03 | 0.0224 | 0.0205 | |
| Noise_ARL+CEMBackbone=ResNet-202025.03 | 50.85 | 0.0271 | 0.0253 | |
| Dropout+CEMBackbone=ResNet-202025.03 | 50.75 | 0.0186 | 0.0172 | |
| DropoutBackbone=ResNet-202025.03 | 50.72 | 0.0163 | 0.0151 |