modules.conv.unet*¶
Function ยท nn.Module
net = unet*(
order=2, kernel_size=3, in_planes=1, out_planes=1
)
Instant presents of mdnc.module.conv.UNet*d.
Arguments¶
Requries
| Argument | Type | Description |
|---|---|---|
order | int | The order of the convolutional layers, could be 1, 2, or 3. |
kernel_size | int or(int,) | The kernel size of each convolutional layer. |
in_planes | int | The channel number of the input data. |
out_planes | int | The channel number of the output data. |
APIs¶
| API | Net depth | Channel | Layer configs | Source |
|---|---|---|---|---|
unet12 | 3 | 64 | [2, 2, 2] | |
unet16 | 4 | 64 | [2, 2, 2, 2] | |
unet17 | 3 | 64 | [3, 3, 3] | |
unet23 | 4 | 64 | [3, 3, 3, 3] | |
unet29 | 5 | 64 | [3, 3, 3, 3, 3] |
where unet29 is a nearly replicated work of milesial/Pytorch-UNet. The only difference of unet29 is two extra convolutional layers for the input and output mapping.
Last update: March 14, 2021