Table 1 Detailed information of the generative network
From: Ancient mural restoration based on a modified generative adversarial network
Name | Type | Kernel | Dilation (η) | Stride | Outputs |
|---|---|---|---|---|---|
Conv1 | conv | 5 × 5 | 1 | 1 × 1 | 64 |
Conv2 | conv | 3 × 3 | 1 | 2 × 2 | 128 |
Conv3 | conv | 3 × 3 | 1 | 1 × 1 | 128 |
Conv4 | conv | 2 × 2 | 1 | 1 × 1 | 256 |
Conv5 | conv | 1 × 1 | 1 | 1 × 1 | 128 |
Conv6 | conv | 1 × 1 | 1 | 1 × 1 | 256 |
Conv7 | conv | 3 × 3 | 1 | 2 × 2 | 128 |
Conv8 | conv | 3 × 3 | 1 | 1 × 1 | 256 |
Dilated conv1 | dilated conv | 3 × 3 | 2 | 1 × 1 | 256 |
Dilated conv2 | dilated conv | 3 × 3 | 4 | 1 × 1 | 256 |
Dilated conv3 | dilated conv | 3 × 3 | 8 | 1 × 1 | 256 |
Conv9 | conv | 3 × 3 | 1 | 1 × 1 | 256 |
Conv10 | conv | 4 × 4 | 1 | 1/2 × 1/2 | 128 |
Deconv1 | deconv | 3 × 3 | 1 | 1 × 1 | 128 |
Conv11 | conv | 4 × 4 | 1 | 1/2 × 1/2 | 64 |
Deconv2 | deconv | 3 × 3 | 1 | 1 × 1 | 32 |
Output | output | 3 × 3 | 1 | 1 × 1 | 3 |