J3：DenseNet算法实战与解析

• 🍨 本文为🔗365天深度学习训练营 中的学习记录博客
• 🍖 原作者：K同学啊 | 接辅导、项目定制
• 🚀 文章来源：K同学的学习圈子

前言

主要介绍DenseNet模型，它的基本思路与ResNet一致,但是它建立的是前面所有层与后面层的密集连接，通过特征图在channel上的连接来实现特征重用

一、模型介绍

相比ResNet, DenseNet提出了一个更为激进的密集连接机制: 即互相连接所有的层，具体来说就是每个层都会接受其前面所有层作为其额外的输入。
下图为ResNet网络的短路连接机制

下图为DenseNet网络的短路连接机制

二、网络结构

1.DenseNet网络结构

DenseNet网络中使用DenseBlock + Transition的结构，其中DenseBlock是包含很多层的模块,每个层的特征图大小相同，层与层之间采用密集连接方式。而Transition层是两个相邻的DenseBlock,并且通过pooling使特征图大小降低。

2. DenseBlock + Transition结构

3. DenseBlock 非线性结构

在DenseBlock中, 各个层的特征图大小一致, 可以在channel维度上连接，DenseBlock基本结构是BN + ReLU +(33)Conv的结构，如下图所示

由于后面层的输入会非常大, DenseBlock内部可以采用bottleneck层来减少计算量, 主要是原有的结构增加11的Conv, 即BN + ReLU + 11Conv + BN + ReLU +33Conv, 称为DenseBlock 结构

三、代码实现

1.导入相关包

import torch
import torch.nn as nn
import torchvision.transforms as transforms
import torchvision
from torchvision import datasets, transforms
import torch.nn.functional as F
from torchsummary import summary
import os,PIL,pathlib,warnings
from  collections import OrderedDict

2. DenseBlock 内部结构

class DenseLayer(nn.Module):
    def __init__(self, in_channels, growth_rate):
        super(DenseLayer, self).__init__()
        self.conv = nn.Conv2d(in_channels, growth_rate, kernel_size=3, padding=1, bias=False)
        self.bn = nn.BatchNorm2d(growth_rate)
        self.relu = nn.ReLU(inplace=True)

    def forward(self, x):
        out = self.conv(x)
        out = self.bn(out)
        out = self.relu(out)
        out = torch.cat((x, out), dim=1)
        return out

3.DenseBlock

class DenseBlock(nn.Module):
    def __init__(self, in_channels, growth_rate, num_layers):
        super(DenseBlock, self).__init__()
        self.layers = nn.ModuleList([DenseLayer(in_channels + i * growth_rate, growth_rate) for i in range(num_layers)])

    def forward(self, x):
        for layer in self.layers:
            x = layer(x)
        return x

4. Transition 层

class TransitionLayer(nn.Module):
    def __init__(self, in_channels, out_channels):
        super(TransitionLayer, self).__init__()
        self.conv = nn.Conv2d(in_channels, out_channels, kernel_size=1, bias=False)
        self.bn = nn.BatchNorm2d(out_channels)
        self.relu = nn.ReLU(inplace=True)
        self.avgpool = nn.AvgPool2d(kernel_size=2, stride=2)

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

5.DenseNet网路

class DenseNet(nn.Module):
    def __init__(self, growth_rate=32, block_config=(6, 12, 24, 16), num_classes=10):
        super(DenseNet, self).__init__()

        # initial convolutional layer
        self.features = nn.Sequential(
            nn.Conv2d(3, 64, kernel_size=3, stride=1, padding=1, bias=False),
            nn.BatchNorm2d(64),
            nn.ReLU(inplace=True)
        )

        # dense blocks and transition layers
        in_channels = 64
        for i, num_layers in enumerate(block_config):
            block = DenseBlock(in_channels, growth_rate, num_layers)
            self.features.add_module(f'denseblock{i + 1}', block)
            in_channels = in_channels + num_layers * growth_rate
            if i != len(block_config) - 1:
                trans = TransitionLayer(in_channels, in_channels // 2)
                self.features.add_module(f'transition{i + 1}', trans)
                in_channels = in_channels // 2

        # final batch normalization and classification layer
        self.features.add_module('bn', nn.BatchNorm2d(in_channels))
        self.features.add_module('relu', nn.ReLU(inplace=True))
        self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
        self.classifier = nn.Linear(in_channels, num_classes)

    def forward(self, x):
        features = self.features(x)
        out = self.avgpool(features)
        out = out.view(features.size(0), -1)
        out = self.classifier(out)
        return out

6.网络模型

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = DenseNet().to(device)
summary(model, (3, 224, 224))

四、在CIFAR10上训练

1.加载CIFAR10

import torchvision.datasets as datasets
import torchvision.transforms as transforms

# 定义预处理转换
transform = transforms.Compose([
    transforms.ToTensor(),  # 将 PIL 图像转换为张量
    transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))  # 归一化
])

# 加载 CIFAR-10 数据集
train_dataset = datasets.CIFAR10(root='./data', train=True, download=True, transform=transform)
test_dataset = datasets.CIFAR10(root='./data', train=False, download=True, transform=transform)

# 创建数据加载器
train_loader = torch.utils.data.DataLoader(train_dataset, batch_size=64, shuffle=True)
test_loader = torch.utils.data.DataLoader(test_dataset, batch_size=64, shuffle=False)

2.训练和测试函数

from tqdm import tqdm
def train (dataloader, model,loss_fn,optimizer):
    size = len(dataloader.dataset)
    num_batches = len(dataloader)
    train_loss = 0
    train_acc = 0
    par = tqdm(dataloader)
    for x,y in par:
        x,y = x.to(device),y.to(device)
        pred = model(x)
        loss = loss_fn(pred,y)
        train_loss += loss.item()
        train_acc += (pred.argmax(1) == y).type(torch.float).sum().item()
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()
        par.set_description(desc= f'loss={loss.item():.4f}')
    train_acc /= size
    train_loss /= num_batches
    return train_acc, train_loss
def test (dataloader, model,loss_fn):
    size = len(dataloader.dataset)
    num_batches = len(dataloader)
    test_loss = 0
    test_acc = 0
    with torch.no_grad():
        par = tqdm(dataloader)
        for x,y in par:
            x,y = x.to(device),y.to(device)
            pred = model(x)
            test_loss += loss_fn(pred,y).item()
            test_acc += (pred.argmax(1) == y).type(torch.float).sum().item()
            par.set_description(desc= f'loss={test_loss:.4f}')
    test_acc /= size
    test_loss /= num_batches
    return test_acc, test_loss

3.正式训练

loss_fn = nn.CrossEntropyLoss()
epochs = 50
train_loss = []
train_acc = []
test_loss = []
test_acc = []
best_acc = 0
learning_rate = 1e-3
optimizer = torch.optim.Adam(model.parameters(),lr=learning_rate)
for epoch in range(1,epochs+1):
    model.train()
    epoch_train_acc,epoch_train_loss = train(train_loader,model,loss_fn,optimizer)
    model.eval()
    epoch_test_acc,epoch_test_loss = test(test_loader,model,loss_fn)
    train_loss.append(epoch_train_loss)
    train_acc.append(epoch_train_acc)
    test_loss.append(epoch_test_loss)
    test_acc.append(epoch_test_acc)
    lr = optimizer.param_groups[0]['lr']
    tmp = "Epoch:{},lr:{:.6f},train_loss:{:.4f},train_acc:{:.4f},test_loss:{:.4f},test_acc:{:.4f}"
    print(tmp.format(epoch,lr,epoch_train_loss,epoch_train_acc,epoch_test_loss,epoch_test_acc))

五、结果可视化

import matplotlib.pyplot as plt
import warnings
warnings.filterwarnings("ignore")               #忽略警告信息
plt.rcParams['font.sans-serif']    = ['SimHei'] # 用来正常显示中文标签
plt.rcParams['axes.unicode_minus'] = False      # 用来正常显示负号
plt.rcParams['figure.dpi']         = 600        #分辨率

epochs_range = range(epochs)

plt.figure(figsize=(12, 3))
plt.subplot(1, 2, 1)

plt.plot(epochs_range, train_acc, label='Training Accuracy')
plt.plot(epochs_range, test_acc, label='Test Accuracy')
plt.legend(loc='lower right')
plt.title('Training and Validation Accuracy')

plt.subplot(1, 2, 2)
plt.plot(epochs_range, train_loss, label='Training Loss')
plt.plot(epochs_range, test_loss, label='Test Loss')
plt.legend(loc='upper right')
plt.title('Training and Validation Loss')
plt.show()