深度学习P5-运动鞋识别

  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

---

我的环境：

1.语言：python3.7

2.编译器：pycharm

---

一、前期准备

1、设置GPU

import torch
import torch.nn as nn
import torchvision.transforms as transforms
import torchvision
from torchvision import transforms, datasets

import os,PIL,pathlib
from torch.optim.lr_scheduler import StepLR, MultiStepLR, LambdaLR, ExponentialLR, CosineAnnealingLR, ReduceLROnPlateau

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

device

device(type='cuda') 

2、导入数据 

import os,PIL,random,pathlib

data_dir = "E:\TF环境\运动鞋"
data_dir = pathlib.Path(data_dir)

data_paths = list(data_dir.glob('*'))
classeNames = [str(path).split("\\")[3] for path in data_paths]
classeNames

1. 使用pathlib.Path()函数将字符串类型的文件夹路径转换为pathlib.Path对象，以便后续操作。

2. 使用glob('*')方法获取了data_dir目录下所有文件和文件夹的路径。'*'通配符代表匹配任意文件或文件夹名。

3. 使用列表推导式遍历data_paths列表中的每个路径，通过将路径转换为字符串并以''为分隔符，然后取第二个元素作为类别名称，并将其添加到classNames列表中。

['test','train'] 

# 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.RandomHorizontalFlip(), # 随机水平翻转
    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。
])

train_dataset = datasets.ImageFolder("E:\TF环境\运动鞋\train\",transform=train_transforms)
test_dataset  = datasets.ImageFolder("E:\TF环境\运动鞋\test\",transform=train_transforms)

transforms.Compose()用于构建数据预处理的操作序列。它可以将多个transform操作组合在一起，以便在数据加载和训练过程中应用这些操作。 

1. transforms.Resize([224, 224])：将图像大小调整为指定的尺寸，这里是将图像的宽度和高度分别调整为224。

2. transforms.ToTensor()：将图像转换为张量形式，将像素值从0-255缩放到0-1之间。

3. transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])：对图像进行归一化处理，通过减去均值(mean)再除以标准差(std)，以使得图像在各个通道上的数值分布接近于标准正态分布。这里给出的均值和标准差是用于ImageNet数据集训练的经验值。

train_dataset.class_to_idx

batch_size = 32

train_dl = torch.utils.data.DataLoader(train_dataset,
                                           batch_size=batch_size,
                                           shuffle=True,
                                           num_workers=1)
test_dl = torch.utils.data.DataLoader(test_dataset,
                                          batch_size=batch_size,
                                          shuffle=True,
                                          num_workers=1)

batch_size = 32 表示每个批次加载的样本数量为32个，即每次训练或测试的时候都会同时处理32个样本。

train_dl = torch.utils.data.DataLoader(train_dataset, batch_size=batch_size, shuffle=True, num_workers=1) 创建了一个训练集的数据加载器。train_dataset是训练集的数据集对象，batch_size指定每个批次加载的样本数量，shuffle=True表示在每个epoch（整个训练集迭代一次）之前将训练集打乱顺序，num_workers=1表示使用一个线程来加载数据。

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

Shape of X [N, C, H, W]:  torch.Size([32, 3, 224, 224])
Shape of y:  torch.Size([32]) torch.int64

二、构建CNN网络

  对于一般的CNN网络来说，都是由特征提取网络和分类网络构成，其中特征提取网络用于提取图片的特征，分类网络用于将图片进行分类。

1. torch.nn.Conv2d()详
 
函数原型：
torch.nn.Conv2d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros', device=None, dtype=None)
 
关键参数说明：
- in_channels ( int ) – 输入图像中的通道数
- out_channels ( int ) – 卷积产生的通道数
- kernel_size ( int or tuple ) – 卷积核的大小
- stride ( int or tuple , optional ) -- 卷积的步幅。默认值：1
- padding ( int , tuple或str , optional ) – 添加到输入的所有四个边的填充。默认值：0
- padding_mode (字符串,可选) – 'zeros', 'reflect', 'replicate'或'circular'. 默认：'zeros'

2. torch.nn.Linear()详解
 
函数原型：
>torch.nn.Linear(in_features, out_features, bias=True, device=None, dtype=None)
 
关键参数说明：
 
- in_features：每个输入样本的大小
- out_features：每个输出样本的大小

3. torch.nn.MaxPool2d()详解
 
函数原型：
>torch.nn.MaxPool2d(kernel_size, stride=None, padding=0, dilation=1, return_indices=False, ceil_mode=False)
 
关键参数说明：
 
- kernel_size：最大的窗口大小
- stride：窗口的步幅，默认值为`kernel_size`
- padding：填充值，默认为0
- dilation：控制窗口中元素步幅的参数

import torch.nn.functional as F

class Model(nn.Module):
    def __init__(self):
        super(Model, self).__init__()
        self.conv1=nn.Sequential(
            nn.Conv2d(3, 12, kernel_size=5, padding=0), # 12*220*220
            nn.BatchNorm2d(12),
            nn.ReLU())
        
        self.conv2=nn.Sequential(
            nn.Conv2d(12, 12, kernel_size=5, padding=0), # 12*216*216
            nn.BatchNorm2d(12),
            nn.ReLU())
        
        self.pool3=nn.Sequential(
            nn.MaxPool2d(2))                              # 12*108*108
        
        self.conv4=nn.Sequential(
            nn.Conv2d(12, 24, kernel_size=5, padding=0), # 24*104*104
            nn.BatchNorm2d(24),
            nn.ReLU())
        
        self.conv5=nn.Sequential(
            nn.Conv2d(24, 24, kernel_size=5, padding=0), # 24*100*100
            nn.BatchNorm2d(24),
            nn.ReLU())
        
        self.pool6=nn.Sequential(
            nn.MaxPool2d(2))                              # 24*50*50

        self.dropout = nn.Sequential(
            nn.Dropout(0.2))
        
        self.fc=nn.Sequential(
            nn.Linear(24*50*50, len(classeNames)))
        
    def forward(self, x):
        
        batch_size = x.size(0)
        x = self.conv1(x)  # 卷积-BN-激活
        x = self.conv2(x)  # 卷积-BN-激活
        x = self.pool3(x)  # 池化
        x = self.conv4(x)  # 卷积-BN-激活
        x = self.conv5(x)  # 卷积-BN-激活
        x = self.pool6(x)  # 池化
        x = self.dropout(x)
        x = x.view(batch_size, -1)  # flatten 变成全连接网络需要的输入 (batch, 24*50*50) ==> (batch, -1), -1 此处自动算出的是24*50*50
        x = self.fc(x)
       
        return x

device = "cuda" if torch.cuda.is_available() else "cpu"
print("Using {} device".format(device))

model = Model().to(device)
model

  三、模型训练 

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 imgs, target in dataloader:
            imgs, target = imgs.to(device), target.to(device)
            
            # 计算loss
            target_pred = model(imgs)
            loss        = loss_fn(target_pred, target)
            
            test_loss += loss.item()
            test_acc  += (target_pred.argmax(1) == target).type(torch.float).sum().item()

    test_acc  /= size
    test_loss /= num_batches

    return test_acc, test_loss

 3、设置动态学习率

def adjust_learning_rate(optimizer, epoch, start_lr):
    # 每 2 个epoch衰减到原来的 0.92
    lr = start_lr * （0.92 ** (epoch // 2)）
    for param_group in optimizer.param_groups:
        param_group['lr'] = lr

learn_rate = 1e-4 # 初始学习率
optimizer  = torch.optim.SGD(model.parameters(), lr=learn_rate)

  函数adjust_learning_rate(optimizer, epoch, start_lr)接受三个参数：optimizer表示优化器对象，epoch表示当前的训练轮数，start_lr表示初始学习率。

  这段代码中使用了指数衰减的方法来调整学习率。每两个epoch，学习率会衰减到原来的0.92倍。具体实现是通过将初始学习率start_lr乘以一个衰减系数(0.92 ** (epoch // 2))获得新的学习率lr。然后，遍历优化器中的所有参数组，将它们的学习率设置为新的学习率lr。

  最后，我们通过使用torch.optim.SGD来创建一个SGD优化器optimizer，并将模型中的参数和学习率传入作为参数。这样就完成了学习率的调整。

调用官方动态学习率接口

 lambda1 = lambda epoch: (0.92 ** (epoch // 2)
 optimizer = torch.optim.SGD(model.parameters(), lr=learn_rate)
 scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer, lr_lambda=lambda1) #选定调整方法

4、正式训练 

loss_fn    = nn.CrossEntropyLoss() # 创建损失函数
epochs     = 40

train_loss = []
train_acc  = []
test_loss  = []
test_acc   = []

for epoch in range(epochs):
    # 更新学习率（使用自定义学习率时使用）
    adjust_learning_rate(optimizer, epoch, learn_rate)
    
    model.train()
    epoch_train_acc, epoch_train_loss = train(train_dl, model, loss_fn, optimizer)
    # scheduler.step() # 更新学习率（调用官方动态学习率接口时使用）
    
    model.eval()
    epoch_test_acc, epoch_test_loss = test(test_dl, model, loss_fn)
    
    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))
print('Done')

 Epoch: 1, Train_acc:53.2%, Train_loss:0.767, Test_acc:53.9%, Test_loss:0.675, Lr:1.00E-04
Epoch: 2, Train_acc:61.4%, Train_loss:0.660, Test_acc:60.5%, Test_loss:0.632, Lr:9.20E-05
Epoch: 3, Train_acc:70.5%, Train_loss:0.600, Test_acc:72.4%, Test_loss:0.570, Lr:9.20E-05
Epoch: 4, Train_acc:70.3%, Train_loss:0.582, Test_acc:71.1%, Test_loss:0.545, Lr:8.46E-05
Epoch: 5, Train_acc:71.9%, Train_loss:0.557, Test_acc:69.7%, Test_loss:0.541, Lr:8.46E-05
Epoch: 6, Train_acc:76.1%, Train_loss:0.529, Test_acc:72.4%, Test_loss:0.526, Lr:7.79E-05
Epoch: 7, Train_acc:75.5%, Train_loss:0.488, Test_acc:73.7%, Test_loss:0.484, Lr:7.79E-05
Epoch: 8, Train_acc:78.3%, Train_loss:0.493, Test_acc:75.0%, Test_loss:0.509, Lr:7.16E-05
Epoch: 9, Train_acc:79.1%, Train_loss:0.454, Test_acc:75.0%, Test_loss:0.511, Lr:7.16E-05
Epoch:10, Train_acc:84.5%, Train_loss:0.428, Test_acc:75.0%, Test_loss:0.507, Lr:6.59E-05
Epoch:11, Train_acc:82.7%, Train_loss:0.433, Test_acc:76.3%, Test_loss:0.451, Lr:6.59E-05
Epoch:12, Train_acc:86.5%, Train_loss:0.397, Test_acc:75.0%, Test_loss:0.484, Lr:6.06E-05
Epoch:13, Train_acc:85.9%, Train_loss:0.404, Test_acc:76.3%, Test_loss:0.434, Lr:6.06E-05
Epoch:14, Train_acc:88.8%, Train_loss:0.375, Test_acc:76.3%, Test_loss:0.524, Lr:5.58E-05
Epoch:15, Train_acc:88.4%, Train_loss:0.376, Test_acc:78.9%, Test_loss:0.456, Lr:5.58E-05
Epoch:16, Train_acc:89.0%, Train_loss:0.366, Test_acc:76.3%, Test_loss:0.469, Lr:5.13E-05
Epoch:17, Train_acc:87.8%, Train_loss:0.369, Test_acc:76.3%, Test_loss:0.492, Lr:5.13E-05
Epoch:18, Train_acc:89.0%, Train_loss:0.348, Test_acc:75.0%, Test_loss:0.436, Lr:4.72E-05
Epoch:19, Train_acc:90.6%, Train_loss:0.331, Test_acc:75.0%, Test_loss:0.461, Lr:4.72E-05
Epoch:20, Train_acc:89.8%, Train_loss:0.340, Test_acc:75.0%, Test_loss:0.463, Lr:4.34E-05
Epoch:21, Train_acc:93.0%, Train_loss:0.318, Test_acc:76.3%, Test_loss:0.440, Lr:4.34E-05
Epoch:22, Train_acc:92.0%, Train_loss:0.314, Test_acc:77.6%, Test_loss:0.462, Lr:4.00E-05
Epoch:23, Train_acc:89.8%, Train_loss:0.321, Test_acc:77.6%, Test_loss:0.422, Lr:4.00E-05
Epoch:24, Train_acc:92.6%, Train_loss:0.313, Test_acc:77.6%, Test_loss:0.459, Lr:3.68E-05
Epoch:25, Train_acc:92.2%, Train_loss:0.310, Test_acc:77.6%, Test_loss:0.427, Lr:3.68E-05
Epoch:26, Train_acc:91.8%, Train_loss:0.305, Test_acc:76.3%, Test_loss:0.409, Lr:3.38E-05
Epoch:27, Train_acc:93.2%, Train_loss:0.304, Test_acc:77.6%, Test_loss:0.436, Lr:3.38E-05
Epoch:28, Train_acc:91.8%, Train_loss:0.310, Test_acc:78.9%, Test_loss:0.445, Lr:3.11E-05
Epoch:29, Train_acc:92.4%, Train_loss:0.300, Test_acc:77.6%, Test_loss:0.426, Lr:3.11E-05
Epoch:30, Train_acc:94.2%, Train_loss:0.285, Test_acc:76.3%, Test_loss:0.423, Lr:2.86E-05
Epoch:31, Train_acc:92.2%, Train_loss:0.285, Test_acc:77.6%, Test_loss:0.419, Lr:2.86E-05
Epoch:32, Train_acc:93.8%, Train_loss:0.284, Test_acc:77.6%, Test_loss:0.401, Lr:2.63E-05
Epoch:33, Train_acc:92.6%, Train_loss:0.287, Test_acc:77.6%, Test_loss:0.390, Lr:2.63E-05
Epoch:34, Train_acc:94.6%, Train_loss:0.267, Test_acc:76.3%, Test_loss:0.423, Lr:2.42E-05
Epoch:35, Train_acc:94.2%, Train_loss:0.269, Test_acc:76.3%, Test_loss:0.437, Lr:2.42E-05
Epoch:36, Train_acc:93.6%, Train_loss:0.268, Test_acc:77.6%, Test_loss:0.425, Lr:2.23E-05
Epoch:37, Train_acc:94.2%, Train_loss:0.275, Test_acc:76.3%, Test_loss:0.414, Lr:2.23E-05
Epoch:38, Train_acc:94.6%, Train_loss:0.267, Test_acc:76.3%, Test_loss:0.404, Lr:2.05E-05
Epoch:39, Train_acc:95.4%, Train_loss:0.256, Test_acc:78.9%, Test_loss:0.411, Lr:2.05E-05
Epoch:40, Train_acc:94.2%, Train_loss:0.270, Test_acc:78.9%, Test_loss:0.419, Lr:1.89E-05

 四、结果可视化

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

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()