# Source: https://github.com/doiken23/DeepLab_pytorch
"""
DeepLab v2 - DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs <https://arxiv.org/abs/1606.00915>`_
"""
import math
import torch.nn as nn
import torch.nn.functional as F
import torchvision.models as models
model_url = 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth'
__all__ = ['DeepLabv2_ASPP', 'DeepLabv2_FOV']
class Atrous_Bottleneck(nn.Module):
expansion = 4
def __init__(self, inplanes, planes, stride=1, rate=1, downsample=None):
super(Atrous_Bottleneck, self).__init__()
self.conv1 = nn.Conv2d(inplanes, planes, kernel_size=1, bias=False)
self.bn1 = nn.BatchNorm2d(planes)
self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=stride,
dilation=rate, padding=rate, bias=False)
self.bn2 = nn.BatchNorm2d(planes)
self.conv3 = nn.Conv2d(planes, planes * 4, kernel_size=1, bias=False)
self.bn3 = nn.BatchNorm2d(planes * 4)
self.relu = nn.ReLU(inplace=True)
self.downsample = downsample
self.stride = stride
def forward(self, x):
residual = x
out = self.conv1(x)
out = self.bn1(out)
out = self.relu(out)
out = self.conv2(out)
out = self.bn2(out)
out = self.relu(out)
out = self.conv3(out)
out = self.bn3(out)
if self.downsample is not None:
residual = self.downsample(x)
out += residual
out = self.relu(out)
return out
class Atrous_ResNet_features(nn.Module):
def __init__(self, block, layers, pretrained=False):
super(Atrous_ResNet_features, self).__init__()
self.inplanes = 64
self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3,
bias=False)
self.bn1 = nn.BatchNorm2d(64)
self.relu = nn.ReLU(inplace=True)
self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
self.layer1 = self._make_layer(block, 64, layers[0], stride=1, rate=1)
self.layer2 = self._make_layer(block, 128, layers[1], stride=2, rate=1)
self.layer3 = self._make_layer(block, 256, layers[2], stride=1, rate=2)
self.layer4 = self._make_layer(block, 512, layers[3], stride=1, rate=4)
for m in self.modules():
if isinstance(m, nn.Conv2d):
n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
m.weight.data.normal_(0, math.sqrt(2. / n))
elif isinstance(m, nn.BatchNorm2d):
m.weight.data.fill_(1)
m.bias.data.zero_()
if pretrained:
print('load the pre-trained model.')
resnet = models.resnet101(pretrained)
self.conv1 = resnet.conv1
self.bn1 = resnet.bn1
self.layer1 = resnet.layer1
self.layer2 = resnet.layer2
def _make_layer(self, block, planes, blocks, stride=1, rate=1):
downsample = None
if stride != 1 or self.inplanes != planes * block.expansion:
downsample = nn.Sequential(
nn.Conv2d(self.inplanes, planes * block.expansion,
kernel_size=1, stride=stride, bias=False),
nn.BatchNorm2d(planes * block.expansion),
)
layers = []
layers.append(block(self.inplanes, planes, stride, rate, downsample))
self.inplanes = planes * block.expansion
for i in range(1, blocks):
layers.append(block(self.inplanes, planes, stride=1, rate=rate))
return nn.Sequential(*layers)
def forward(self, x):
x = self.conv1(x)
x = self.bn1(x)
x = self.relu(x)
x = self.maxpool(x)
x = self.layer1(x)
x = self.layer2(x)
x = self.layer3(x)
x = self.layer4(x)
return x
class Atrous_module(nn.Module):
def __init__(self, inplanes, num_classes, rate):
super(Atrous_module, self).__init__()
planes = inplanes
self.atrous_convolution = nn.Conv2d(inplanes, planes, kernel_size=3,
stride=1, padding=rate, dilation=rate)
self.fc1 = nn.Conv2d(planes, planes, kernel_size=1, stride=1)
self.fc2 = nn.Conv2d(planes, num_classes, kernel_size=1, stride=1)
def forward(self, x):
x = self.atrous_convolution(x)
x = self.fc1(x)
x = self.fc2(x)
return x
[docs]class DeepLabv2_ASPP(nn.Module):
"""
DeeplabV2 Resnet implementation with ASPP.
"""
def __init__(self, num_classes, small=True, pretrained=False, **kwargs):
super(DeepLabv2_ASPP, self).__init__()
block = Atrous_Bottleneck
self.resnet_features = Atrous_ResNet_features(block, [3, 4, 23, 3], pretrained)
if small:
rates = [2, 4, 8, 12]
else:
rates = [6, 12, 18, 24]
self.aspp1 = Atrous_module(2048, num_classes, rate=rates[0])
self.aspp2 = Atrous_module(2048, num_classes, rate=rates[1])
self.aspp3 = Atrous_module(2048, num_classes, rate=rates[2])
self.aspp4 = Atrous_module(2048, num_classes, rate=rates[3])
def forward(self, x):
x = self.resnet_features(x)
x1 = self.aspp1(x)
x2 = self.aspp2(x)
x3 = self.aspp3(x)
x4 = self.aspp4(x)
x = x1 + x2 + x3 + x4
x = F.interpolate(x, scale_factor=8, mode='bilinear')
return x
[docs]class DeepLabv2_FOV(nn.Module):
"""
DeeplabV2 Resnet implementation with FOV.
"""
def __init__(self, num_classes, pretrained=True, **kwargs):
super(DeepLabv2_FOV, self).__init__()
block = Atrous_Bottleneck
self.resnet_features = Atrous_ResNet_features(block, [3, 4, 23, 3], pretrained)
self.atrous = Atrous_module(2048, num_classes, rate=12)
def forward(self, x):
x = self.resnet_features(x)
x = self.atrous(x)
x = F.interpolate(x, scale_factor=8, mode='bilinear')
return x