Image Classification and Reconstruction Defense on CIFAR100 (val test)
68.43AccuracyBottleneck
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| BottleneckBackbone=VGG112025.03 | 68.43 | 0.0007 | 0.0007 | 0.0009 | 0.0008 | |
| Bottleneck+CEMBackbone=VGG112025.03 | 68.42 | 0.0058 | 0.0059 | 0.0019 | 0.002 | |
| ResSFL+CEMBackbone=VGG112025.03 | 66.96 | 0.0093 | 0.0095 | 0.0043 | 0.0046 | |
| ResSFLBackbone=VGG112025.03 | 66.94 | 0.0051 | 0.0052 | 0.0019 | 0.002 | |
| DistCorr+CEMBackbone=VGG112025.03 | 66.27 | 0.0074 | 0.0075 | 0.0044 | 0.0045 | |
| DistCorrBackbone=VGG112025.03 | 66.21 | 0.0054 | 0.0055 | 0.0023 | 0.0025 | |
| Dropout+CEMBackbone=VGG112025.03 | 65.92 | 0.006 | 0.0061 | 0.0031 | 0.0032 | |
| DropoutBackbone=VGG112025.03 | 65.85 | 0.0042 | 0.0043 | 0.0016 | 0.0018 | |
| Average w/o CEMBackbone=VGG112025.03 | 65.81 | 0.0117 | 0.012 | 0.0063 | 0.0073 | |
| Average w/ CEMBackbone=VGG112025.03 | 65.75 | 0.0165 | 0.0168 | 0.0095 | 0.0102 | |
| Noise_NopeekBackbone=VGG112025.03 | 65.55 | 0.0063 | 0.0063 | 0.0037 | 0.0038 | |
| Noise_Nopeek+CEMBackbone=VGG112025.03 | 65.33 | 0.0083 | 0.0082 | 0.0052 | 0.0052 | |
| PATROLBackbone=VGG112025.03 | 65.1 | 0.0335 | 0.0346 | 0.0196 | 0.0238 | |
| PATROL+CEMBackbone=VGG112025.03 | 65.07 | 0.0419 | 0.0423 | 0.0235 | 0.0253 | |
| Noise_ARLBackbone=VGG112025.03 | 62.58 | 0.027 | 0.0277 | 0.0143 | 0.0164 | |
| Noise_ARL+CEMBackbone=VGG112025.03 | 62.34 | 0.0373 | 0.038 | 0.0219 | 0.0236 |