• 🍨 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

学习目的：

1. 跑通RestNet34代码
2. 根据官方代码的结构输出+代码结构图，手动搭建RestNet34算法网络

一、 前期准备

关于环境

• 语言环境：Python3.13
• 编译器：vsCode
• 深度学习环境：torch==2.11.0+cu130；torchvision==0.26.0+cu130torchvision==0.26.0+cu130

1.设置GPU

设置分析环境：

import torch
import torch.nn as nn
import torchvision.transforms as transforms
import torchvision
from torchvision import transforms, datasets
import os,PIL,pathlib,warnings

warnings.filterwarnings("ignore")             #忽略警告信息

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
device

运行结果：

2. 导入数据

由于本次是本地数据，因此不需要有download代码，只需读取路径里的数据集即可

import os,PIL,random,pathlib

data_dir = './data/'
data_dir = pathlib.Path(data_dir)

data_paths  = list(data_dir.glob('*'))
classeNames = [str(path).split("\\")[1] for path in data_paths]

# 关于transforms.Compose的更多介绍可以参考：https://blog.csdn.net/qq_38251616/article/details/124878863
train_transforms = transforms.Compose([
    transforms.Resize([224, 224]),  # 将输入图片resize成统一尺寸
    transforms.ToTensor(),          # 将PIL Image或numpy.ndarray转换为tensor，并归一化到[0,1]之间
    transforms.Normalize(           # 标准化处理-->转换为标准正太分布（高斯分布），使模型更容易收敛
        mean=[0.485, 0.456, 0.406], 
        std=[0.229, 0.224, 0.225])  # 其中 mean=[0.485,0.456,0.406]与std=[0.229,0.224,0.225] 从数据集中随机抽样计算得到的。
])

test_transform = transforms.Compose([
    transforms.Resize([224, 224]),  # 将输入图片resize成统一尺寸
    transforms.ToTensor(),          # 将PIL Image或numpy.ndarray转换为tensor，并归一化到[0,1]之间
    transforms.Normalize(           # 标准化处理-->转换为标准正太分布（高斯分布），使模型更容易收敛
        mean=[0.485, 0.456, 0.406], 
        std=[0.229, 0.224, 0.225])  # 其中 mean=[0.485,0.456,0.406]与std=[0.229,0.224,0.225] 从数据集中随机抽样计算得到的。
])

total_data = datasets.ImageFolder("./data/",transform=train_transforms)
total_data

运行结果：

total_data.class_to_idx

运行结果：

3.划分数据集

train_size = int(0.8 * len(total_data))
test_size  = len(total_data) - train_size
train_dataset, test_dataset = torch.utils.data.random_split(total_data, [train_size, test_size])
train_dataset, test_dataset

运行结果：

batch_size = 64

train_dl = torch.utils.data.DataLoader(train_dataset,
                                       batch_size=batch_size,
                                       shuffle=True,
                                       num_workers=4)
test_dl = torch.utils.data.DataLoader(test_dataset,
                                      batch_size=batch_size,
                                      shuffle=False,
                                      num_workers=4)

for X, y in test_dl:
    print("Shape of X [N, C, H, W]: ", X.shape)
    print("Shape of y: ", y.shape, y.dtype)
    break

运行结果：

查看一个批次的数据形状：X 是图像张量，形状为 [64, 3, 224, 224]（64 张，3 通道，高宽 224）；y 是标签，形状为 [64]，类型为 torch.int64。

二、ResNet34模型

1. 调用模型

from torchvision.models import resnet34

# 加载预训练模型，并且对模型进行微调
model = resnet34(pretrained = True).to(device) # 加载预训练的resnet34模型

for param in model.parameters():
    param.requires_grad = False # 冻结模型的参数，这样子在训练的时候只训练最后一层的参数

# 修改模型的fc层，即(fc): Linear(in_features=512, out_features=2, bias=True)
# 注意查看我们下方打印出来的模型
model.fc = nn.Linear(512,len(classeNames)) # 修改vgg16模型中最后一层全连接层，输出目标类别个数
model.to(device)  
model

运行结果：

# 统计模型参数量以及其他指标
import torchsummary as summary
summary.summary(model, (3, 224, 224))

运行结果：

2.模型拆解

结合deepseek及CSDN内相关推文进行拆解

import torch.nn.functional as F
 
#定义残差网络
class BasicBlock(nn.Module):
    # 判断残差结构中，主分支的卷积核个数是否发生变化，不变则为1
    expansion = 1
    def __init__(self,in_channel,out_channel,stride=1,downsample=None):
        super(BasicBlock,self).__init__()
        #卷积层1
        self.conv1 = nn.Conv2d(in_channels=in_channel,out_channels=out_channel,
                               kernel_size=3,stride=stride,padding=1,bias=False
        )
        # 使用批量归一化
        self.bn1 = nn.BatchNorm2d(out_channel)
        self.relu = nn.ReLU()
         #卷积层2
        self.conv2 = nn.Conv2d(in_channels=out_channel,out_channels=out_channel,
                               kernel_size=3,stride=1,padding=1,bias=False)
        self.bn2 = nn.BatchNorm2d(out_channel)
        self.downsample =downsample
    def forward(self,x):
 
        identity = x
        if self.downsample is not None:
            identity=self.downsample(x)  
        out = self.conv1(x)
        out =self.bn1(out)
        out = self.relu(out)
 
        out = self.conv2(out)
        out = self.bn2(out)
        out +=identity
        out = self.relu(out)
        return out

#定义ResNet类
class ResNet(nn.Module):
    def__init__(self,block,blocks_num,num_classes=1000):
        super(ResNet,self).__init__()
        # maxpool的输出通道数为64，残差结构的输入通道为64
        self.in_channel = 64
        self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3)   # 64*112*112
        self.bn1 = nn.BatchNorm2d(64)                                       # 64*112*112
        self.relu = nn.ReLU(inplace=True)
        self.Pool1 = nn.MaxPool2d(kernel_size=(3, 3), stride=(2, 2))        # 64*56*56
        # layer1: # 64*56*56   (不做降采样，stride=1)
        self.layer1 = self._make_layer(block, 64,blocks_num[0])
        # layer2: # 128*28*28   (降采样，stride=2)
        self.layer2 = self._make_layer(block, 128, blocks_num[1],stride=2)
        # layer3: # 256*14*14   (降采样，stride=2)
        self.layer3 = self._make_layer(block, 256, blocks_num[2],stride=2)
        # layer4: # 512*7*7   (降采样，stride=2)
        self.layer4 = self._make_layer(block, 512, blocks_num[3],stride=2)
        # 自适应全局平均池化：无论输入多大，都压缩成 1x1
        self.avgpool = nn.AdaptiveAvgPool2d((1,1)) 
        # 全连接层：512 通道 -> 1000 类别
        self.fc = nn.Linear(512*block.expansion,num_classes)
        
    def _make_layer(self, block, out_channels, blocks, stride=1):
        downsample = None
        # 1. 判断是否需要降采样（尺寸变小 或 通道数变化）
        # 注意：BasicBlock.expansion = 1，所以 out_channels * 1 = out_channels
        if stride != 1 or self.in_channels != out_channels * block.expansion:
            # 构建捷径上的 1x1 卷积层（用来强行对齐输入和输出的形状）
            downsample = nn.Sequential(
                nn.Conv2d(self.in_channels, out_channels * block.expansion,
                          kernel_size=1, stride=stride, bias=False),
                nn.BatchNorm2d(out_channels * block.expansion),
            )

        # 2. 构建本阶段的层列表
        layers = []
        # 第一个残差块：负责处理降采样（可能带 downsample）
        layers.append(block(self.in_channels, out_channels, stride, downsample))

        # 更新当前通道数，供后面的块使用
        self.in_channels = out_channels * block.expansion

        # 3. 剩下的 blocks-1 个残差块（步长均为1，尺寸和通道都不变）
        for _ in range(1, blocks):
            layers.append(block(self.in_channels, out_channels, stride=1, downsample=None))

        # 返回一个顺序容器
        return nn.Sequential(*layers)

    def forward(self, x):
        # 预处理
        x = self.conv1(x)
        x = self.bn1(x)
        x = self.relu(x)
        x = self.maxpool(x)

        # 4 个阶段
        x = self.layer1(x)
        x = self.layer2(x)
        x = self.layer3(x)
        x = self.layer4(x)

        # 分类头
        x = self.avgpool(x)
        x = torch.flatten(x, 1)  # 展平 (batch_size, 512, 1, 1) -> (batch_size, 512)
        x = self.fc(x)

        return x

三、 训练模型

1. 编写训练函数

# 训练循环
def train(dataloader, model, loss_fn, optimizer):
    size = len(dataloader.dataset)  # 训练集的大小
    num_batches = len(dataloader)   # 批次数目, (size/batch_size，向上取整)

    train_loss, train_acc = 0, 0  # 初始化训练损失和正确率
    
    for X, y in dataloader:  # 获取图片及其标签
        X, y = X.to(device), y.to(device)
        
        # 计算预测误差
        pred = model(X)          # 网络输出
        loss = loss_fn(pred, y)  # 计算网络输出和真实值之间的差距，targets为真实值，计算二者差值即为损失
        
        # 反向传播
        optimizer.zero_grad()  # grad属性归零
        loss.backward()        # 反向传播
        optimizer.step()       # 每一步自动更新
        
        # 记录acc与loss
        train_acc  += (pred.argmax(1) == y).type(torch.float).sum().item()
        train_loss += loss.item()
            
    train_acc  /= size
    train_loss /= num_batches

    return train_acc, train_loss

2. 编写测试函数

测试函数和训练函数大致相同，但是由于不进行梯度下降对网络权重进行更新，所以不需要传入优化器

def test (dataloader, model, loss_fn):
    size        = len(dataloader.dataset)  # 测试集的大小
    num_batches = len(dataloader)          # 批次数目, (size/batch_size，向上取整)
    test_loss, test_acc = 0, 0
    
    # 当不进行训练时，停止梯度更新，节省计算内存消耗
    with torch.no_grad():
        for X, y in dataloader:
            X, y = X.to(device), y.to(device)
            
            # 计算loss
            y_pred = model(X)
            loss   = loss_fn(y_pred, y)
            
            test_loss += loss.item()
            test_acc  += (y_pred.argmax(1) == y).type(torch.float).sum().item()

    test_acc  /= size
    test_loss /= num_batches

    return test_acc, test_loss

3. 正式训练

import copy

optimizer  = torch.optim.Adam(model.parameters(), lr= 1e-4)
loss_fn    = nn.CrossEntropyLoss() # 创建损失函数

epochs     = 10
train_loss = []
train_acc  = []
test_loss  = []
test_acc   = []

best_acc = 0    # 设置一个最佳准确率，作为最佳模型的判别指标

for epoch in range(epochs):
    
    model.train()
    epoch_train_acc, epoch_train_loss = train(train_dl, model, loss_fn, optimizer)
    
    model.eval()
    epoch_test_acc, epoch_test_loss = test(test_dl, model, loss_fn)
    
    # 保存最佳模型到 best_model
    if epoch_test_acc > best_acc:
        best_acc   = epoch_test_acc
        best_model = copy.deepcopy(model)
    
    train_acc.append(epoch_train_acc)
    train_loss.append(epoch_train_loss)
    test_acc.append(epoch_test_acc)
    test_loss.append(epoch_test_loss)
    
    # 获取当前的学习率
    lr = optimizer.state_dict()['param_groups'][0]['lr']
    
    template = ('Epoch:{:2d}, Train_acc:{:.1f}%, Train_loss:{:.3f}, Test_acc:{:.1f}%, Test_loss:{:.3f}, Lr:{:.2E}')
    print(template.format(epoch+1, epoch_train_acc*100, epoch_train_loss, 
                          epoch_test_acc*100, epoch_test_loss, lr))
    
# 保存最佳模型到文件中
PATH = './best_model.pth'  # 保存的参数文件名
torch.save(best_model.state_dict(), PATH)

print('Done')

运行结果：

四、 结果可视化

1. Loss与Accuracy图

import matplotlib.pyplot as plt
#隐藏警告
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']         = 100        #分辨率

from datetime import datetime
current_time = datetime.now() # 获取当前时间

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.xlabel(current_time) # 打卡请带上时间戳，否则代码截图无效

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()

运行结果：

2. 模型评估

best_model.eval()
epoch_test_acc, epoch_test_loss = test(test_dl, best_model, loss_fn)

epoch_test_acc, epoch_test_loss

运行结果：

五、感想

本周主要学习的是ResNat残差网络。ResNet 通过残差连接，让输入可以绕过若干卷积层后直接与输出相加，从而缓解深层网络训练困难的问题，解决了模型无法加深的问题。本周重点学习了残差网络的主要结构和构建逻辑，了解了下采样的意义，但是主要代码书写还要依靠AI，有待提高。