deeplabv3 pytorch 代码

以下是使用PyTorch实现的DeepLabv3代码：

import torch
import torch.nn as nn
import torch.nn.functional as F


class ASPP(nn.Module):
    def __init__(self, in_channels, out_channels=256, rates=[6, 12, 18]):
        super(ASPP, self).__init__()
        self.conv1 = nn.Conv2d(in_channels, out_channels, kernel_size=1)
        self.conv2 = nn.Conv2d(in_channels, out_channels, kernel_size=3, padding=rates[0], dilation=rates[0])
        self.conv3 = nn.Conv2d(in_channels, out_channels, kernel_size=3, padding=rates[1], dilation=rates[1])
        self.conv4 = nn.Conv2d(in_channels, out_channels, kernel_size=3, padding=rates[2], dilation=rates[2])
        self.conv5 = nn.Conv2d(out_channels * 5, out_channels, kernel_size=1)
        self.bn = nn.BatchNorm2d(out_channels)
        self.relu = nn.ReLU()

    def forward(self, x):
        x1 = self.conv1(x)
        x2 = self.conv2(x)
        x3 = self.conv3(x)
        x4 = self.conv4(x)
        x5 = F.avg_pool2d(x, kernel_size=x.size()[2:])
        x5 = self.conv1(x5)
        x5 = F.interpolate(x5, size=x4.size()[2:], mode='bilinear', align_corners=True)
        x = torch.cat([x1, x2, x3, x4, x5], dim=1)
        x = self.conv5(x)
        x = self.bn(x)
        x = self.relu(x)
        return x


class DeepLabV3(nn.Module):
    def __init__(self, num_classes=21):
        super(DeepLabV3, self).__init__()
        self.resnet = torchvision.models.resnet101(pretrained=True)
        self.aspp = ASPP(in_channels=2048, out_channels=256)
        self.conv1 = nn.Conv2d(256, 48, kernel_size=1)
        self.conv2 = nn.Conv2d(304, 256, kernel_size=3, padding=1)
        self.conv3 = nn.Conv2d(256, 256, kernel_size=3, padding=1)
        self.conv4 = nn.Conv2d(256, num_classes, kernel_size=1)
        self.relu = nn.ReLU()
        self.dropout = nn.Dropout(p=0.5)
        self.upsample = nn.Upsample(scale_factor=8, mode='bilinear', align_corners=True)

    def forward(self, x):
        x = self.resnet.conv1(x)
        x = self.resnet.bn1(x)
        x = self.resnet.relu(x)
        x = self.resnet.maxpool(x)
        x = self.resnet.layer1(x)
        x = self.resnet.layer2(x)
        x = self.resnet.layer3(x)
        x = self.resnet.layer4(x)
        x = self.aspp(x)
        x = self.conv1(x)
        x = self.dropout(x)
        x = F.interpolate(x, size=(x.size()[2]*2, x.size()[3]*2), mode='bilinear', align_corners=True)
        x = torch.cat([x, self.resnet.layer2[-1]], dim=1)
        x = self.conv2(x)
        x = self.dropout(x)
        x = F.interpolate(x, size=(x.size()[2]*2, x.size()[3]*2), mode='bilinear', align_corners=True)
        x = torch.cat([x, self.resnet.layer1[-1]], dim=1)
        x = self.conv3(x)
        x = self.dropout(x)
        x = self.conv4(x)
        x = self.upsample(x)
        return x

其中，ASPP模块实现了空洞卷积池化模块，DeepLabV3模型则基于ASPP模块实现了完整的DeepLabv3网络。