Awesome
Results on CIFAR100
The table prvoides the models and results of various models on CIFAR100. Learning rate =0.1 and will be divided by 10 every 70 epochs. Total 300 epochs. Using SGD optimizer, momentum=0.9, weight_decay=5e-4. Loss is CrossEntropyLoss. Batch-size=512.
| Model | Parameters | Flops | CIFAR-100 |
|---|---|---|---|
| PreActResNet18 | - | - | 74.91% |
| PreActResNet50 | - | - | 77.39% |
| PreActResNet101 | - | - | 77.74% |
| SEResNet18 | - | - | 75.19% |
| SEResNet50 | - | - | 77.91% |
| SEResNet101 | - | - | 78.03% |
| PSEResNet18 | - | - | 74.97% |
| PSEResNet50 | - | - | 77.45% |
| PSEResNet101 | - | - | 77.88% |
| CPSEResNet18 | - | - | 75.25% |
| CPSEResNet50 | - | - | 77.43% |
| CPSEResNet101 | - | - | 77.61% |
| SPPSEResNet18 | - | - | 75.41% |
| SPPSEResNet50 | - | 78.21% | |
| SPPSEResNet101 | - | - | 78.11 |
| PSPPSEResNet18 | - | - | 75.01% |
| PSPPSEResNet50 | - | - | 78.11% |
| PSPPSEResNet101 | - | - | 78.35% |
| CPSPPSEResNet18 | - | - | 75.56% |
| CPSPPSEResNet50 | - | - | 77.95% |
| CPSPPSEResNet101 | - | - | 79.17% |
For a better understanding, we reschedule the table as follows:
| Model | 18-Layer | 50-Layer | 101-Layer |
|---|---|---|---|
| PreActResNet | 74.91% | 77.39% | 77.74% |
| SEResNet | 75.19% | 77.91% | 78.03% |
| PSEResNet | 74.97% | 77.45% | 77.88% |
| CPSEResNet | 75.25% | 77.43% | 77.61% |
| SPPSEResNet | 75.41% | 78.21% | 78.11% |
| PSPPSEResNet | 75.01% | 78.11% | 78.35% |
| CPSPPSEResNet | 75.56% | 77.95% | 79.17% |