ImageNet pretrained model layer dimensions #30
Description
I'm studying this code and would like to know if my understanding is correct.
Generator:
| layer | gifn | gain_ifn | bias_ifn |
|---|---|---|---|
| 1 | (100, 128 * 8 * 4 * 4) | 128 * 8 * 4 * 4 | 128 * 8 * 4 * 4 |
| 2 | (1024, 512, 5, 5) | 512 | 512 |
| 3 | (512, 256, 5, 5) | 256 | 256 |
| 4 | (256, 128, 5, 5) | 128 | 128 |
| 5 | (128, 64, 5, 5) | 64 | 64 |
| 6 | (64, 32, 5, 5) | 32 | 32 |
| output | (32, 3, 5, 5) | ---- | ---- |
Discriminator:
| layer | gifn | gain_ifn | bias_ifn |
|---|---|---|---|
| 1 | (32, 3, 5, 5) | ---- | ---- |
| 2 | (64, 32, 5, 5) | 64 | 64 |
| 3 | (128, 64, 5, 5) | 128 | 128 |
| 4 | (256, 128, 5, 5) | 256 | 256 |
| 5 | (512, 256, 5, 5) | 512 | 512 |
| 6 | (1024, 512, 5, 5) | 1024 | 1024 |
| output | (128 * 8 * 4 * 4, 1) | ---- | ---- |
Are these dimensions correct for the ImageNet model?