### 使用PyTorch实现的语义分割开源模型 #### 1. DeepLab系列模型 DeepLab是一种广泛应用于语义分割任务的经典模型。它通过引入空间金字塔池化模块（ASPP）以及空洞卷积来增强特征提取能力[^2]。以下是基于PyTorch实现的一个简单版本： ```python import torch.nn as nn from torchvision.models.segmentation import deeplabv3_resnet50 class CustomDeeplabV3(nn.Module): def __init__(self, num_classes=21): super(CustomDeeplabV3, self).__init__() self.model = deeplabv3_resnet50(pretrained=True) self.model.classifier[-1] = nn.Conv2d(256, num_classes, kernel_size=1) def forward(self, x): return self.model(x)['out'] ``` 此代码片段展示了如何利用`torchvision`库中的预训练权重快速搭建一个支持自定义类别的DeepLabV3模型。 --- #### 2. U-Net架构 U-Net因其结构紧凑且性能优越而被广泛应用在医学图像处理领域。其核心思想是对称编码器-解码器设计，辅以跳跃连接提升细节恢复能力[^4]。下面提供了一个基础版的U-Net实现： ```python import torch.nn.functional as F import torch.nn as nn def double_conv(in_channels, out_channels): return nn.Sequential( nn.Conv2d(in_channels, out_channels, kernel_size=3), nn.ReLU(inplace=True), nn.Conv2d(out_channels, out_channels, kernel_size=3), nn.ReLU(inplace=True) ) class UNet(nn.Module): def __init__(self, n_class): super().__init__() self.dconv_down1 = double_conv(3, 64) self.dconv_down2 = double_conv(64, 128) self.dconv_down3 = double_conv(128, 256) self.dconv_down4 = double_conv(256, 512) self.maxpool = nn.MaxPool2d(2) self.up = nn.Upsample(scale_factor=2, mode='bilinear', align_corners=True) self.dconv_up3 = double_conv(256 + 512, 256) self.dconv_up2 = double_conv(128 + 256, 128) self.dconv_up1 = double_conv(128 + 64, 64) self.conv_last = nn.Conv2d(64, n_class, 1) def forward(self, x): conv1 = self.dconv_down1(x) x = self.maxpool(conv1) conv2 = self.dconv_down2(x) x = self.maxpool(conv2) conv3 = self.dconv_down3(x) x = self.maxpool(conv3) x = self.dconv_down4(x) x = self.up(x) x = torch.cat([x, conv3], dim=1) x = self.dconv_up3(x) x = self.up(x) x = torch.cat([x, conv2], dim=1) x = self.dconv_up2(x) x = self.up(x) x = torch.cat([x, conv1], dim=1) x = self.dconv_up1(x) out = self.conv_last(x) return out ``` 该网络实现了标准的下采样路径和上采样路径，并通过双线性插值完成尺寸匹配[^3]。 --- #### 3. FCN（全卷积神经网络） FCN作为早期经典的语义分割方法之一，主要特点是将传统CNN的最后一层全连接替换为卷积操作，从而保留输入的空间维度信息[^1]。以下是一个简单的FCN-8s实现： ```python import torch.nn as nn from torchvision.models import vgg16 class FCNs(nn.Module): def __init__(self, pretrained_net, n_class): super().__init__() self.n_class = n_class self.pretrained_net = pretrained_net self.relu = nn.ReLU(inplace=True) # 定义其他必要的组件... def forward(self, x): output = self.pretrained_net.features(x) score_pool3 = ... # 提取中间层特征并进行融合 score_pool4 = ... score_fuse = ... return score_fuse ``` 由于篇幅限制，未完全展示具体逻辑，但提供了基本思路供参考。 --- #### 总结 以上分别介绍了三种主流的语义分割模型及其对应的PyTorch实现方式。每种模型都有各自的特点与适用范围，在实际应用中可根据数据集规模、硬件资源等因素灵活选择。