AI_python_yoooger/Ai_tottle/map/uav_seg/models/deeplabv3plus.py

import torch
import torch.nn as nn
import torch.nn.functional as F
from torchvision.models import resnet50, ResNet50_Weights

class ASPP(nn.Module):
    """空洞空间金字塔池化模块"""
    def __init__(self, in_channels, out_channels, rates):
        super(ASPP, self).__init__()
        
        self.aspp = nn.ModuleList()
        
        # 1x1 卷积
        self.aspp.append(nn.Sequential(
            nn.Conv2d(in_channels, out_channels, 1, bias=False),
            nn.BatchNorm2d(out_channels),
            nn.ReLU(inplace=True)
        ))
        
        # 空洞卷积
        for rate in rates:
            self.aspp.append(nn.Sequential(
                nn.Conv2d(in_channels, out_channels, 3, padding=rate, dilation=rate, bias=False),
                nn.BatchNorm2d(out_channels),
                nn.ReLU(inplace=True)
            ))
        
        # 全局平均池化
        self.global_avg_pool = nn.Sequential(
            nn.AdaptiveAvgPool2d(1),
            nn.Conv2d(in_channels, out_channels, 1, bias=False),
            nn.BatchNorm2d(out_channels),
            nn.ReLU(inplace=True)
        )
        
        # 输出层
        self.output = nn.Sequential(
            nn.Conv2d(out_channels * (len(rates) + 2), out_channels, 1, bias=False),
            nn.BatchNorm2d(out_channels),
            nn.ReLU(inplace=True),
            nn.Dropout(0.5)
        )
    
    def forward(self, x):
        size = x.size()
        
        outputs = []
        for module in self.aspp:
            outputs.append(module(x))
        
        # 全局平均池化
        global_feat = self.global_avg_pool(x)
        global_feat = F.interpolate(global_feat, size=(size[2], size[3]), mode='bilinear', align_corners=True)
        outputs.append(global_feat)
        
        # 拼接所有特征
        x = torch.cat(outputs, dim=1)
        
        return self.output(x)

class DeepLabV3Plus(nn.Module):
    """DeepLabV3+模型，适用于大尺度遥感图像语义分割"""
    def __init__(self, num_classes, backbone='resnet50', output_stride=16, pretrained=True):
        super(DeepLabV3Plus, self).__init__()
        
        # 设置空洞卷积率
        if output_stride == 16:
            rates = [6, 12, 18]
        elif output_stride == 8:
            rates = [12, 24, 36]
        else:
            raise ValueError("output_stride 必须是 8 或 16!")
        
        # 主干网络
        if backbone == 'resnet50':
            if pretrained:
                self.backbone = resnet50(weights=ResNet50_Weights.DEFAULT)
            else:
                self.backbone = resnet50(weights=None)
            
            self.low_level_features = self.backbone.layer1
            self.high_level_features = nn.Sequential(
                self.backbone.layer2,
                self.backbone.layer3,
                self.backbone.layer4
            )
            low_level_channels = 256
            high_level_channels = 2048
        else:
            raise ValueError(f"不支持的主干网络: {backbone}")
        
        # ASPP模块
        self.aspp = ASPP(high_level_channels, 256, rates)
        
        # 低层特征处理
        self.low_level_conv = nn.Sequential(
            nn.Conv2d(low_level_channels, 48, 1, bias=False),
            nn.BatchNorm2d(48),
            nn.ReLU(inplace=True)
        )
        
        # 解码器
        self.decoder = nn.Sequential(
            nn.Conv2d(304, 256, 3, padding=1, bias=False),
            nn.BatchNorm2d(256),
            nn.ReLU(inplace=True),
            nn.Dropout(0.5),
            nn.Conv2d(256, 256, 3, padding=1, bias=False),
            nn.BatchNorm2d(256),
            nn.ReLU(inplace=True),
            nn.Dropout(0.1),
            nn.Conv2d(256, num_classes, 1)
        )
        
        # 初始化权重
        self._init_weight()
    
    def forward(self, x):
        size = x.size()
        
        # 提取特征
        x = self.backbone.conv1(x)
        x = self.backbone.bn1(x)
        x = self.backbone.relu(x)
        x = self.backbone.maxpool(x)
        
        # 低层特征
        low_level_feat = self.low_level_features(x)
        
        # 高层特征
        x = self.high_level_features(low_level_feat)
        x = self.aspp(x)
        
        # 上采样高层特征
        x = F.interpolate(x, size=low_level_feat.size()[2:], mode='bilinear', align_corners=True)
        
        # 处理低层特征
        low_level_feat = self.low_level_conv(low_level_feat)
        
        # 拼接特征
        x = torch.cat((x, low_level_feat), dim=1)
        
        # 解码
        x = self.decoder(x)
        
        # 上采样到原始大小
        x = F.interpolate(x, size=size[2:], mode='bilinear', align_corners=True)
        
        return x
    
    def _init_weight(self):
        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal_(m.weight)
            elif isinstance(m, nn.BatchNorm2d):
                m.weight.data.fill_(1)
                m.bias.data.zero_()
    
    def get_backbone_params(self):
        modules = [self.backbone.conv1, self.backbone.bn1, self.backbone.layer1, self.backbone.layer2, self.backbone.layer3, self.backbone.layer4]
        for i in range(len(modules)):
            for m in modules[i].named_modules():
                if isinstance(m[1], nn.Conv2d) or isinstance(m[1], nn.BatchNorm2d):
                    for p in m[1].parameters():
                        if p.requires_grad:
                            yield p
    
    def get_decoder_params(self):
        modules = [self.aspp, self.low_level_conv, self.decoder]
        for i in range(len(modules)):
            for m in modules[i].named_modules():
                if isinstance(m[1], nn.Conv2d) or isinstance(m[1], nn.BatchNorm2d):
                    for p in m[1].parameters():
                        if p.requires_grad:
                            yield p

    def freeze_backbone(self):
        for param in self.backbone.parameters():
            param.requires_grad = False

    def unfreeze_backbone(self):
        for param in self.backbone.parameters():
            param.requires_grad = True