Image Classification on CIFAR-10 standard (test) (Anat and Arob Metrics)
84.21Anat AccuracyResNet-50
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ResNet-50Backbone=ResNet-50, Size [MB]=89.7, FPS=162021.10 | 84.21 | 53.05 | |
| WRN-28-4Backbone=WideResNet-28-4, Size [MB]=22.3, FPS=172021.10 | 84 | 51.8 | |
| ResNet-50 + GDWSBackbone=ResNet-50, beta=0.001, Size [MB]=81.9, FPS=372021.10 | 83.72 | 52.94 | |
| WRN-28-4 + GDWSBackbone=WideResNet-28-4, beta=1e-5, Size [MB]=18.9, FPS=652021.10 | 83.27 | 51.7 | |
| RobNetSize [MB]=20.8, FPS=52021.10 | 82.72 | 52.23 | |
| ResNet-18Backbone=ResNet-18, Size [MB]=42.6, FPS=282021.10 | 82.41 | 51.55 | |
| ResNet-18 + GDWSBackbone=ResNet-18, beta=0.005, Size [MB]=29.1, FPS=1042021.10 | 81.17 | 50.98 | |
| VGG-16Backbone=VGG-16, Size [MB]=56.2, FPS=362021.10 | 77.49 | 48.92 | |
| VGG-16 + GDWSBackbone=VGG-16, beta=0.25, Size [MB]=28.7, FPS=1292021.10 | 77.17 | 49.56 |