Image Classification on ImageNet1k (val) (Top-1 Accuracy and ECE)
74.02Top-1 AccuracyPolyLoss
Evaluation results
| Method | Links | ||
|---|---|---|---|
| PolyLossBackbone=ResNet-34, Loss Function=PolyLoss2023.04 | 74.02 | 7.882 | |
| Label Smoothing (LS)Backbone=ResNet-34, Loss Function=Label Smoothing2023.04 | 73.69 | 3.994 | |
| Entropy Regularization (ER)Backbone=ResNet-34, Loss Function=CE with entropy regularization2023.04 | 73.68 | 3.72 | |
| Standard Cross Entropy (CE) LossBackbone=ResNet-34, Loss Function=CE2023.04 | 73.56 | 5.301 | |
| L1 Norm-based regularizationBackbone=ResNet-34, Loss Function=CE with L1 Norm-based regularization2023.04 | 73.21 | 2.625 | |
| Focal Loss (FL)Backbone=ResNet-34, Loss Function=Focal Loss2023.04 | 72.82 | 4.901 |