基于pytorch的DenseNet 11*11图像大小分类代码

以下是使用Pytorch实现基于DenseNet的11 * 11图像分类的示例代码：

import torch
import torch.nn as nn
import torch.optim as optim
from torchvision.datasets import CIFAR10
from torchvision.transforms import transforms
from torch.utils.data import DataLoader
from torch.optim.lr_scheduler import MultiStepLR

# Define hyperparameters
num_epochs = 50
batch_size = 128
learning_rate = 0.01
num_classes = 10
num_layers_per_block = 4
growth_rate = 32
compression_factor = 0.5

# Define data transforms
transform = transforms.Compose([
    transforms.RandomCrop(11),
    transforms.RandomHorizontalFlip(),
    transforms.ToTensor(),
    transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])

# Load CIFAR10 dataset
train_dataset = CIFAR10(root='./data', train=True, download=True, transform=transform)
test_dataset = CIFAR10(root='./data', train=False, download=True, transform=transform)
train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True)
test_loader = DataLoader(test_dataset, batch_size=batch_size, shuffle=False)

# Define DenseNet block
class DenseNetBlock(nn.Module):
    def __init__(self, in_channels, growth_rate, num_layers):
        super(DenseNetBlock, self).__init__()
        self.layers = nn.ModuleList()
        for i in range(num_layers):
            self.layers.append(nn.Sequential(
                nn.BatchNorm2d(in_channels + i * growth_rate),
                nn.ReLU(inplace=True),
                nn.Conv2d(in_channels + i * growth_rate, growth_rate, kernel_size=1, bias=False),
                nn.BatchNorm2d(growth_rate),
                nn.ReLU(inplace=True),
                nn.Conv2d(growth_rate, growth_rate, kernel_size=3, padding=1, bias=False)
            ))
    
    def forward(self, x):
        for layer in self.layers:
            out = layer(x)
            x = torch.cat([x, out], 1)
        return x

# Define DenseNet model
class DenseNet(nn.Module):
    def __init__(self, num_classes, num_layers_per_block, growth_rate, compression_factor):
        super(DenseNet, self).__init__()
        self.features = nn.Sequential(
            nn.Conv2d(3, 2*growth_rate, kernel_size=3, padding=1, bias=False),
            nn.BatchNorm2d(2*growth_rate),
            nn.ReLU(inplace=True),
            nn.MaxPool2d(kernel_size=2, stride=2)
        )
        num_channels = 2 * growth_rate
        self.num_blocks = len(num_layers_per_block)
        for i, num_layers in enumerate(num_layers_per_block):
            block = DenseNetBlock(num_channels, growth_rate, num_layers)
            self.features.add_module("denseblock%d" % (i + 1), block)
            num_channels += num_layers * growth_rate
            if i != self.num_blocks - 1:
                num_channels = int(num_channels * compression_factor)
                trans = nn.Sequential(
                    nn.Conv2d(num_channels, num_channels, kernel_size=1, bias=False),
                    nn.BatchNorm2d(num_channels),
                    nn.ReLU(inplace=True),
                    nn.AvgPool2d(kernel_size=2, stride=2)
                )
                self.features.add_module("transition%d" % (i + 1), trans)
        self.classifier = nn.Linear(num_channels, num_classes)
    
    def forward(self, x):
        x = self.features(x)
        x = nn.functional.avg_pool2d(x, kernel_size=(x.size(2), x.size(3)))
        x = x.view(x.size(0), -1)
        x = self.classifier(x)
        return x

# Initialize model and optimizer
model = DenseNet(num_classes, num_layers_per_block, growth_rate, compression_factor)
optimizer = optim.SGD(model.parameters(), lr=learning_rate, momentum=0.9, weight_decay=1e-4)
scheduler = MultiStepLR(optimizer, milestones=[int(num_epochs*0.5), int(num_epochs*0.75)], gamma=0.1)

# Train model
model.train()
for epoch in range(num_epochs):
    for i, (images, labels) in enumerate(train_loader):
        # Forward pass
        outputs = model(images)
        loss = nn.functional.cross_entropy(outputs, labels)
        
        # Backward and optimize
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()
    
    # Print training progress
    print('Epoch [{}/{}], Loss: {:.4f}'.format(epoch+1, num_epochs, loss.item()))
    
    # Update learning rate
    scheduler.step()

# Test model
model.eval()
with torch.no_grad():
    correct = 0
    total = 0
    for images, labels in test_loader:
        outputs = model(images)
        _, predicted = torch.max(outputs.data, 1)
        total += labels.size(0)
        correct += (predicted == labels).sum().item()
    
    # Print test accuracy
    print('Test Accuracy of the model on the 10000 test images: {} %'.format(100 * correct / total))

这个示例代码基于DenseNet的思想，使用了一个堆叠的DenseNet块来构建网络。每个块由多个DenseNet层组成，每个层都通过连接前面所有层的方式进行连接，以最大化信息流动。这个代码还包括数据增强、学习率调度器等常用的训练技巧，以帮助提高模型的性能。