Image Classification on SVHN extra train (test)
2.29Global ScoreVGG8B (Global Backprop)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| VGG8B (Global Backprop)Backbone=VGG8B, Parameters=8.9M, Loss Function=Global2019.01 | 2.29 | — | — | — | |
| WideResNet (Global Backprop)Backbone=WideResNet-16-8, Parameters=11M, Loss Function=Global2019.01 | 1.6 | — | — | — | |
| WideResNet+CO (Global Backprop)Backbone=WideResNet-16-8, Parameters=11M, Loss Function=Global, Data Augmentation=Cutout2019.01 | 1.3 | — | — | — | |
| VGG8B (Prediction Loss)Backbone=VGG8B, Parameters=8.9M, Loss Function=Local pred2019.01 | — | 2.12 | — | — | |
| VGG8B (Predsim Loss)Backbone=VGG8B, Parameters=8.9M, Loss Function=Local predsim2019.01 | — | — | — | 1.74 | |
| VGG8B (Similarity Loss)Backbone=VGG8B, Parameters=8.9M, Loss Function=Local sim2019.01 | — | — | 1.89 | — | |
| VGG8B+CO (Predsim Loss)Backbone=VGG8B, Parameters=8.9M, Loss Function=Local predsim, Data Augmentation=Cutout2019.01 | — | — | — | 1.65 |