# Adapted from:
# https://github.com/milesial/Pytorch-UNet/blob/master/unet/unet_parts.py
# By Emma Avetissian, @aemmav
""" Parts of the U-Net model """
import torch
import torch.nn as nn
import torch.nn.functional as F
[docs]
class DoubleConv(nn.Module):
"""(convolution => [BN] => ReLU) * 2"""
def __init__(self, in_channels, out_channels, mid_channels=None):
super().__init__()
if not mid_channels:
mid_channels = out_channels
self.double_conv = nn.Sequential(
nn.Conv3d(in_channels, mid_channels, kernel_size=3, padding=1, bias=False),
nn.BatchNorm3d(mid_channels),
nn.ReLU(inplace=True),
nn.Conv3d(mid_channels, out_channels, kernel_size=3, padding=1, bias=False),
nn.BatchNorm3d(out_channels),
nn.ReLU(inplace=True),
)
[docs]
def forward(self, x):
return self.double_conv(x)
[docs]
class Down(nn.Module):
"""Downscaling with maxpool then double conv"""
def __init__(self, in_channels, out_channels):
super().__init__()
self.maxpool_conv = nn.Sequential(
nn.MaxPool3d(2), DoubleConv(in_channels, out_channels)
)
[docs]
def forward(self, x):
return self.maxpool_conv(x)
[docs]
class Up(nn.Module):
"""Upscaling then double conv"""
def __init__(self, in_channels, out_channels, trilinear=True):
super().__init__()
# if trilinear, use the normal convolutions to reduce the number of channels
if trilinear:
self.up = nn.Upsample(scale_factor=2, mode="trilinear", align_corners=True)
self.conv = DoubleConv(in_channels, out_channels, in_channels // 2)
else:
self.up = nn.ConvTranspose3d(
in_channels, in_channels // 2, kernel_size=2, stride=2
)
self.conv = DoubleConv(in_channels, out_channels)
[docs]
def forward(self, x1, x2):
x1 = self.up(x1)
# input is CHW
diffY = x2.size()[2] - x1.size()[2] # height
diffX = x2.size()[3] - x1.size()[3] # width
diffZ = x2.size()[4] - x1.size()[4] # depth
x1 = F.pad(
x1,
[
diffX // 2,
diffX - diffX // 2,
diffY // 2,
diffY - diffY // 2,
diffZ // 2,
diffZ - diffZ // 2,
],
)
# if you have padding issues, see
# https://github.com/HaiyongJiang/U-Net-Pytorch-Unstructured-Buggy/commit/0e854509c2cea854e247a9c615f175f76fbb2e3a
# https://github.com/xiaopeng-liao/Pytorch-UNet/commit/8ebac70e633bac59fc22bb5195e513d5832fb3bd
x = torch.cat([x2, x1], dim=1)
return self.conv(x)
[docs]
class OutConv(nn.Module):
def __init__(self, in_channels, out_channels):
super(OutConv, self).__init__()
self.conv = nn.Conv3d(in_channels, out_channels, kernel_size=1)
[docs]
def forward(self, x):
return self.conv(x)
[docs]
class UNet_3D(nn.Module):
"""
3D U-Net model.
Adapted from:
https://github.com/milesial/Pytorch-UNet/blob/master/unet/unet_model.py
By Emma Avetissian, @aemmav
Parameters
----------
n_channels : int
The number of input channels.
n_classes : int
The number of output channels.
trilinear : bool
Whether to use trilinear interpolation or not.
"""
def __init__(self, n_channels, n_classes, trilinear=False):
super(UNet_3D, self).__init__()
self.n_channels = n_channels
self.n_classes = n_classes
self.trilinear = trilinear
self.inc = DoubleConv(n_channels, 64)
self.down1 = Down(64, 128)
self.down2 = Down(128, 256)
self.down3 = Down(256, 512)
factor = 2 if trilinear else 1
self.down4 = Down(512, 1024 // factor)
self.up1 = Up(1024, 512 // factor, trilinear)
self.up2 = Up(512, 256 // factor, trilinear)
self.up3 = Up(256, 128 // factor, trilinear)
self.up4 = Up(128, 64, trilinear)
self.outc = OutConv(64, n_classes)
[docs]
def forward(self, x):
x1 = self.inc(x)
x2 = self.down1(x1)
x3 = self.down2(x2)
x4 = self.down3(x3)
x5 = self.down4(x4)
x = self.up1(x5, x4)
x = self.up2(x, x3)
x = self.up3(x, x2)
x = self.up4(x, x1)
logits = self.outc(x)
return logits
