# Copyright 2020 - 2021 MONAI Consortium
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
# http://www.apache.org/licenses/LICENSE-2.0
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from typing import Optional, Sequence, Union
import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
from monai.networks.blocks.segresnet_block import ResBlock, get_conv_layer, get_norm_layer, get_upsample_layer
from monai.networks.layers.factories import Act, Dropout
from monai.utils import UpsampleMode
[docs]class SegResNet(nn.Module):
"""
SegResNet based on `3D MRI brain tumor segmentation using autoencoder regularization
<https://arxiv.org/pdf/1810.11654.pdf>`_.
The module does not include the variational autoencoder (VAE).
The model supports 2D or 3D inputs.
Args:
spatial_dims: spatial dimension of the input data. Defaults to 3.
init_filters: number of output channels for initial convolution layer. Defaults to 8.
in_channels: number of input channels for the network. Defaults to 1.
out_channels: number of output channels for the network. Defaults to 2.
dropout_prob: probability of an element to be zero-ed. Defaults to ``None``.
norm_name: feature normalization type, this module only supports group norm,
batch norm and instance norm. Defaults to ``group``.
num_groups: number of groups to separate the channels into. Defaults to 8.
use_conv_final: if add a final convolution block to output. Defaults to ``True``.
blocks_down: number of down sample blocks in each layer. Defaults to ``[1,2,2,4]``.
blocks_up: number of up sample blocks in each layer. Defaults to ``[1,1,1]``.
upsample_mode: [``"deconv"``, ``"nontrainable"``, ``"pixelshuffle"``]
The mode of upsampling manipulations.
Using the ``nontrainable`` modes cannot guarantee the model's reproducibility. Defaults to``nontrainable``.
- ``deconv``, uses transposed convolution layers.
- ``nontrainable``, uses non-trainable `linear` interpolation.
- ``pixelshuffle``, uses :py:class:`monai.networks.blocks.SubpixelUpsample`.
"""
def __init__(
self,
spatial_dims: int = 3,
init_filters: int = 8,
in_channels: int = 1,
out_channels: int = 2,
dropout_prob: Optional[float] = None,
norm_name: str = "group",
num_groups: int = 8,
use_conv_final: bool = True,
blocks_down: tuple = (1, 2, 2, 4),
blocks_up: tuple = (1, 1, 1),
upsample_mode: Union[UpsampleMode, str] = UpsampleMode.NONTRAINABLE,
):
super().__init__()
if spatial_dims not in (2, 3):
raise AssertionError("spatial_dims can only be 2 or 3.")
self.spatial_dims = spatial_dims
self.init_filters = init_filters
self.blocks_down = blocks_down
self.blocks_up = blocks_up
self.dropout_prob = dropout_prob
self.norm_name = norm_name
self.num_groups = num_groups
self.upsample_mode = UpsampleMode(upsample_mode)
self.use_conv_final = use_conv_final
self.convInit = get_conv_layer(spatial_dims, in_channels, init_filters)
self.down_layers = self._make_down_layers()
self.up_layers, self.up_samples = self._make_up_layers()
self.relu = Act[Act.RELU](inplace=True)
self.conv_final = self._make_final_conv(out_channels)
if dropout_prob is not None:
self.dropout = Dropout[Dropout.DROPOUT, spatial_dims](dropout_prob)
def _make_down_layers(self):
down_layers = nn.ModuleList()
blocks_down, spatial_dims, filters, norm_name, num_groups = (
self.blocks_down,
self.spatial_dims,
self.init_filters,
self.norm_name,
self.num_groups,
)
for i in range(len(blocks_down)):
layer_in_channels = filters * 2 ** i
pre_conv = (
get_conv_layer(spatial_dims, layer_in_channels // 2, layer_in_channels, stride=2)
if i > 0
else nn.Identity()
)
down_layer = nn.Sequential(
pre_conv,
*[
ResBlock(spatial_dims, layer_in_channels, norm_name=norm_name, num_groups=num_groups)
for _ in range(blocks_down[i])
],
)
down_layers.append(down_layer)
return down_layers
def _make_up_layers(self):
up_layers, up_samples = nn.ModuleList(), nn.ModuleList()
upsample_mode, blocks_up, spatial_dims, filters, norm_name, num_groups = (
self.upsample_mode,
self.blocks_up,
self.spatial_dims,
self.init_filters,
self.norm_name,
self.num_groups,
)
n_up = len(blocks_up)
for i in range(n_up):
sample_in_channels = filters * 2 ** (n_up - i)
up_layers.append(
nn.Sequential(
*[
ResBlock(spatial_dims, sample_in_channels // 2, norm_name=norm_name, num_groups=num_groups)
for _ in range(blocks_up[i])
]
)
)
up_samples.append(
nn.Sequential(
*[
get_conv_layer(spatial_dims, sample_in_channels, sample_in_channels // 2, kernel_size=1),
get_upsample_layer(spatial_dims, sample_in_channels // 2, upsample_mode=upsample_mode),
]
)
)
return up_layers, up_samples
def _make_final_conv(self, out_channels: int):
return nn.Sequential(
get_norm_layer(self.spatial_dims, self.init_filters, norm_name=self.norm_name, num_groups=self.num_groups),
self.relu,
get_conv_layer(self.spatial_dims, self.init_filters, out_channels=out_channels, kernel_size=1, bias=True),
)
[docs] def forward(self, x):
x = self.convInit(x)
if self.dropout_prob is not None:
x = self.dropout(x)
down_x = []
for down in self.down_layers:
x = down(x)
down_x.append(x)
down_x.reverse()
for i, (up, upl) in enumerate(zip(self.up_samples, self.up_layers)):
x = up(x) + down_x[i + 1]
x = upl(x)
if self.use_conv_final:
x = self.conv_final(x)
return x
[docs]class SegResNetVAE(SegResNet):
"""
SegResNetVAE based on `3D MRI brain tumor segmentation using autoencoder regularization
<https://arxiv.org/pdf/1810.11654.pdf>`_.
The module contains the variational autoencoder (VAE).
The model supports 2D or 3D inputs.
Args:
spatial_dims: spatial dimension of the input data. Defaults to 3.
init_filters: number of output channels for initial convolution layer. Defaults to 8.
in_channels: number of input channels for the network. Defaults to 1.
out_channels: number of output channels for the network. Defaults to 2.
dropout_prob: probability of an element to be zero-ed. Defaults to ``None``.
norm_name: feature normalization type, this module only supports group norm,
batch norm and instance norm. Defaults to ``group``.
num_groups: number of groups to separate the channels into. Defaults to 8.
use_conv_final: if add a final convolution block to output. Defaults to ``True``.
blocks_down: number of down sample blocks in each layer. Defaults to ``[1,2,2,4]``.
blocks_up: number of up sample blocks in each layer. Defaults to ``[1,1,1]``.
upsample_mode: [``"deconv"``, ``"nontrainable"``, ``"pixelshuffle"``]
The mode of upsampling manipulations.
Using the ``nontrainable`` modes cannot guarantee the model's reproducibility. Defaults to `nontrainable`.
- ``deconv``, uses transposed convolution layers.
- ``nontrainable``, uses non-trainable `linear` interpolation.
- ``pixelshuffle``, uses :py:class:`monai.networks.blocks.SubpixelUpsample`.
use_vae: if use the variational autoencoder (VAE) during training. Defaults to ``False``.
input_image_size: the size of images to input into the network. It is used to
determine the in_features of the fc layer in VAE. When ``use_vae == True``, please
ensure that this parameter is set. Defaults to ``None``.
vae_estimate_std: whether to estimate the standard deviations in VAE. Defaults to ``False``.
vae_default_std: if not to estimate the std, use the default value. Defaults to 0.3.
vae_nz: number of latent variables in VAE. Defaults to 256.
Where, 128 to represent mean, and 128 to represent std.
"""
def __init__(
self,
input_image_size: Sequence[int],
vae_estimate_std: bool = False,
vae_default_std: float = 0.3,
vae_nz: int = 256,
spatial_dims: int = 3,
init_filters: int = 8,
in_channels: int = 1,
out_channels: int = 2,
dropout_prob: Optional[float] = None,
norm_name: str = "group",
num_groups: int = 8,
use_conv_final: bool = True,
blocks_down: tuple = (1, 2, 2, 4),
blocks_up: tuple = (1, 1, 1),
upsample_mode: Union[UpsampleMode, str] = "nontrainable",
):
super(SegResNetVAE, self).__init__(
spatial_dims=spatial_dims,
init_filters=init_filters,
in_channels=in_channels,
out_channels=out_channels,
dropout_prob=dropout_prob,
norm_name=norm_name,
num_groups=num_groups,
use_conv_final=use_conv_final,
blocks_down=blocks_down,
blocks_up=blocks_up,
upsample_mode=upsample_mode,
)
self.input_image_size = input_image_size
self.smallest_filters = 16
zoom = 2 ** (len(self.blocks_down) - 1)
self.fc_insize = [s // (2 * zoom) for s in self.input_image_size]
self.vae_estimate_std = vae_estimate_std
self.vae_default_std = vae_default_std
self.vae_nz = vae_nz
self._prepare_vae_modules()
self.vae_conv_final = self._make_final_conv(in_channels)
def _prepare_vae_modules(self):
zoom = 2 ** (len(self.blocks_down) - 1)
v_filters = self.init_filters * zoom
total_elements = int(self.smallest_filters * np.prod(self.fc_insize))
self.vae_down = nn.Sequential(
get_norm_layer(self.spatial_dims, v_filters, norm_name=self.norm_name, num_groups=self.num_groups),
self.relu,
get_conv_layer(self.spatial_dims, v_filters, self.smallest_filters, stride=2, bias=True),
get_norm_layer(
self.spatial_dims, self.smallest_filters, norm_name=self.norm_name, num_groups=self.num_groups
),
self.relu,
)
self.vae_fc1 = nn.Linear(total_elements, self.vae_nz)
self.vae_fc2 = nn.Linear(total_elements, self.vae_nz)
self.vae_fc3 = nn.Linear(self.vae_nz, total_elements)
self.vae_fc_up_sample = nn.Sequential(
get_conv_layer(self.spatial_dims, self.smallest_filters, v_filters, kernel_size=1),
get_upsample_layer(self.spatial_dims, v_filters, upsample_mode=self.upsample_mode),
get_norm_layer(self.spatial_dims, v_filters, norm_name=self.norm_name, num_groups=self.num_groups),
self.relu,
)
def _get_vae_loss(self, net_input: torch.Tensor, vae_input: torch.Tensor):
"""
Args:
net_input: the original input of the network.
vae_input: the input of VAE module, which is also the output of the network's encoder.
"""
x_vae = self.vae_down(vae_input)
x_vae = x_vae.view(-1, self.vae_fc1.in_features)
z_mean = self.vae_fc1(x_vae)
z_mean_rand = torch.randn_like(z_mean)
z_mean_rand.requires_grad_(False)
if self.vae_estimate_std:
z_sigma = self.vae_fc2(x_vae)
z_sigma = F.softplus(z_sigma)
vae_reg_loss = 0.5 * torch.mean(z_mean ** 2 + z_sigma ** 2 - torch.log(1e-8 + z_sigma ** 2) - 1)
x_vae = z_mean + z_sigma * z_mean_rand
else:
z_sigma = self.vae_default_std
vae_reg_loss = torch.mean(z_mean ** 2)
x_vae = z_mean + z_sigma * z_mean_rand
x_vae = self.vae_fc3(x_vae)
x_vae = self.relu(x_vae)
x_vae = x_vae.view([-1, self.smallest_filters] + self.fc_insize)
x_vae = self.vae_fc_up_sample(x_vae)
for up, upl in zip(self.up_samples, self.up_layers):
x_vae = up(x_vae)
x_vae = upl(x_vae)
x_vae = self.vae_conv_final(x_vae)
vae_mse_loss = F.mse_loss(net_input, x_vae)
vae_loss = vae_reg_loss + vae_mse_loss
return vae_loss
[docs] def forward(self, x):
net_input = x
x = self.convInit(x)
if self.dropout_prob is not None:
x = self.dropout(x)
down_x = []
for down in self.down_layers:
x = down(x)
down_x.append(x)
down_x.reverse()
vae_input = x
for i, (up, upl) in enumerate(zip(self.up_samples, self.up_layers)):
x = up(x) + down_x[i + 1]
x = upl(x)
if self.use_conv_final:
x = self.conv_final(x)
if self.training:
vae_loss = self._get_vae_loss(net_input, vae_input)
return x, vae_loss
return x, None