【论文笔记】图像修复MPRNet:Multi-Stage Progressive Image Restoration 含代码解析

**论文笔记：MPRNet**

**Multi-Stage Progressive Image Restoration**

**简介**

MPRNet是一种用于图像修复的深度学习模型，旨在恢复损坏或模糊的图像。该模型通过多阶段逐进式的方式来处理图像修复任务，利用了自适应卷积神经网络（ACNN）和残差连接（Residual Connection）的优势。

**模型架构**

MPRNet的模型架构如下所示：

1. **输入层**: 输入图像大小为256x256。
2. **多阶段逐进式模块**: 模型通过多个阶段逐进式地处理图像修复任务，每个阶段使用不同的卷积核和残差连接来恢复图像细节。
3. **自适应卷积神经网络（ACNN）**: 每个阶段使用ACNN来学习特定尺寸的图像特征。
4. **残差连接（Residual Connection）**: 每个阶段使用残差连接来将前一阶段的输出与当前阶段的输入相加，减少信息丢失。

**多阶段逐进式模块**

每个阶段都有一个自适应卷积神经网络（ACNN）和一个残差连接。ACNN用于学习特定尺寸的图像特征，而残差连接则用于将前一阶段的输出与当前阶段的输入相加。

import torchimport torch.nn as nnclass MPRNet(nn.Module):
 def __init__(self):
 super(MPRNet, self).__init__()
 self.stage1 = nn.Sequential(
 ACNN(3,64),
 nn.ReLU(),
 nn.MaxPool2d(kernel_size=2)
 )
 self.stage2 = nn.Sequential(
 ACNN(64,128),
 nn.ReLU(),
 nn.MaxPool2d(kernel_size=2)
 )
 self.stage3 = nn.Sequential(
 ACNN(128,256),
 nn.ReLU()
 )

 def forward(self, x):
 out1 = self.stage1(x)
 out2 = self.stage2(out1)
 out3 = self.stage3(out2)
 return out3

class ACNN(nn.Module):
 def __init__(self, in_channels, out_channels):
 super(ACNN, self).__init__()
 self.conv = nn.Conv2d(in_channels, out_channels, kernel_size=3)
 self.relu = nn.ReLU()
 self.maxpool = nn.MaxPool2d(kernel_size=2)

 def forward(self, x):
 out = self.conv(x)
 out = self.relu(out)
 out = self.maxpool(out)
 return out

class ResidualConnection(nn.Module):
 def __init__(self, in_channels, out_channels):
 super(ResidualConnection, self).__init__()
 self.conv = nn.Conv2d(in_channels, out_channels, kernel_size=3)
 self.relu = nn.ReLU()

 def forward(self, x):
 out = self.conv(x)
 out = self.relu(out)
 return out

# 训练model = MPRNet()
criterion = nn.MSELoss()
optimizer = torch.optim.Adam(model.parameters(), lr=0.001)

for epoch in range(10):
 optimizer.zero_grad()
 outputs = model(inputs)
 loss = criterion(outputs, labels)
 loss.backward()
 optimizer.step()

# 测试model.eval()
with torch.no_grad():
 outputs = model(inputs)