Table 7 The influence of different downsampleing pooling with ResNet-18 and compared with MGP-Net, experiments trained without Degradation model on Morph \(\text {II}\) dataset.
From: Multi-grained pooling network for age estimation in degraded low-resolution images
Model | Pooling | Downsamples | Image Size | MAE | CS(5) |
|---|---|---|---|---|---|
ResNet-18 | vanilla | 3 | 64\(\times\)64 | 2.61 | 88.94 |
ResNet-18 | AvgPool | 3 | 64\(\times\)64 | 2.57 | 88.88 |
ResNet-18 | MaxPool | 3 | 64\(\times\)64 | 2.62 | 88.25 |
ResNet-18 | SoftPool | 3 | 64\(\times\)64 | 2.55 | 89.17 |
ResNet-18 | MGPool | 3 | 64\(\times\)64 | 2.52 | 89.22 |
MGP-Net | MGPool | 3 | 64\(\times\)64 | 2.43 | 90.57 |
