第P6周：VGG-16算法-Pytorch实现人脸识别

• 🍨 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

前言

要求：

1. 保存训练过程中的最佳模型权重
2. 调用官方的VGG-16网络框架

拔高：

1. 测试集准确率达到60%（难度有点大，但是这个过程可以学到不少）
2. 手动搭建VGG-16网络框架

实验环境：

• 语言环境：Python3.10
• 编译器：Pycharm
• 深度学习环境：Pytorch
  1.torch == 2.4.0+cu121
  2.torchvision == 0.19.0+cu121

一、前期准备

1.设置GPU

iimport torch
import torch.nn as nn
import torchvision.transforms as transforms
import torchvision
from torchvision import transforms, datasets
import os,PIL,pathlib,warnings
import os,PIL,random,pathlib
from torchvision.models import vgg16
warnings.filterwarnings("ignore")  #忽略警告信息

#-------------设置GPU---------------
# 设置硬件设备，如果有GPU则使用，没有则使用cpu
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(device)

代码输出：

cuda

2.导入数据

输入代码：

#---------------------------导入数据----------------------
data_dir = 'F:/lsydata/48-data/'
data_dir = pathlib.Path(data_dir)

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

代码输出：

['Angelina Jolie', 'Brad Pitt', 'Denzel Washington', 'Hugh Jackman', 'Jennifer Lawrence', 'Johnny Depp', 'Kate Winslet', 'Leonardo DiCaprio', 'Megan Fox', 'Natalie Portman', 'Nicole Kidman', 'Robert Downey Jr', 'Sandra Bullock', 'Scarlett Johansson', 'Tom Cruise', 'Tom Hanks', 'Will Smith']

代码输入：

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。
])

total_data = datasets.ImageFolder("F:/lsydata/48-data/", transform=train_transforms)
print(total_data)

代码输出：

Dataset ImageFolder
    Number of datapoints: 1800
    Root location: F:/lsydata/48-data/
    StandardTransform
Transform: Compose(
               Resize(size=[224, 224], interpolation=bilinear, max_size=None, antialias=True)
               RandomHorizontalFlip(p=0.5)
               ToTensor()
               Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
           )

代码输入：

print(total_data.class_to_idx)

代码输出：

{'Angelina Jolie': 0, 'Brad Pitt': 1, 'Denzel Washington': 2, 'Hugh Jackman': 3, 'Jennifer Lawrence': 4, 'Johnny Depp': 5, 'Kate Winslet': 6, 'Leonardo DiCaprio': 7, 'Megan Fox': 8, 'Natalie Portman': 9, 'Nicole Kidman': 10, 'Robert Downey Jr': 11, 'Sandra Bullock': 12, 'Scarlett Johansson': 13, 'Tom Cruise': 14, 'Tom Hanks': 15, 'Will Smith': 16}

3.划分数据集

代码输入：

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])
print(train_dataset)
print(test_dataset)

代码输出：

<torch.utils.data.dataset.Subset object at 0x000001A52C990910>
<torch.utils.data.dataset.Subset object at 0x000001A52C990820>

代码输入：

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)

if __name__ == '__main__':
    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

代码输出：

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

二、调用官方的VGG-16模型

代码输入：

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

# 加载预训练模型，并且对模型进行微调
model = vgg16(pretrained=True).to(device)  # 加载预训练的vgg16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classifier模块的第6层（即：(6): Linear(in_features=4096, out_features=2, bias=True)）
# 注意查看我们下方打印出来的模型
model.classifier._modules['6'] = nn.Linear(4096, len(classNames))  # 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)
print(model)

代码输出：

Using cuda device
VGG(
  (features): Sequential(
    (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (1): ReLU(inplace=True)
    (2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (3): ReLU(inplace=True)
    (4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (5): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (6): ReLU(inplace=True)
    (7): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (8): ReLU(inplace=True)
    (9): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (10): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (11): ReLU(inplace=True)
    (12): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (13): ReLU(inplace=True)
    (14): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (15): ReLU(inplace=True)
    (16): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (17): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (18): ReLU(inplace=True)
    (19): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (20): ReLU(inplace=True)
    (21): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (22): ReLU(inplace=True)
    (23): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (24): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (25): ReLU(inplace=True)
    (26): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (27): ReLU(inplace=True)
    (28): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (29): ReLU(inplace=True)
    (30): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (avgpool): AdaptiveAvgPool2d(output_size=(7, 7))
  (classifier): Sequential(
    (0): Linear(in_features=25088, out_features=4096, bias=True)
    (1): ReLU(inplace=True)
    (2): Dropout(p=0.5, inplace=False)
    (3): Linear(in_features=4096, out_features=4096, bias=True)
    (4): ReLU(inplace=True)
    (5): Dropout(p=0.5, inplace=False)
    (6): Linear(in_features=4096, out_features=17, bias=True)
  )
)

三、训练模型

1.编写训练函数

#-------------------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，一共60000张图片
    num_batches = len(dataloader) #批次数目，1875（60000/32）

    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

#--------------------编写测试函数-------------------------------
def test (dataloader, model, loss_fn):
    size = len(dataloader.dataset)  #测试集的大小，一共10000张图片
    num_batches = len(dataloader)   #批次数目，313（10000/32=312.5，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

# 调用官方动态学习率接口时使用
learn_rate = 1e-4 # 初始学习率
lambda1 = lambda epoch: 0.92 ** (epoch // 4)
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

#---------------------------正式训练-------------------------
loss_fn = nn.CrossEntropyLoss() # 创建损失函数
epochs = 40

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

if __name__ == '__main__':
    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')

代码输出：

Epoch: 1, Train_acc:7.3%, Train_loss:2.892, Test_acc:8.3%, Test_loss:2.802, Lr:1.00E-04
Epoch: 2, Train_acc:10.0%, Train_loss:2.865, Test_acc:9.7%, Test_loss:2.794, Lr:1.00E-04
Epoch: 3, Train_acc:10.6%, Train_loss:2.835, Test_acc:12.5%, Test_loss:2.762, Lr:1.00E-04
Epoch: 4, Train_acc:12.3%, Train_loss:2.801, Test_acc:12.8%, Test_loss:2.746, Lr:9.20E-05
Epoch: 5, Train_acc:11.0%, Train_loss:2.782, Test_acc:13.6%, Test_loss:2.731, Lr:9.20E-05
Epoch: 6, Train_acc:12.6%, Train_loss:2.767, Test_acc:12.8%, Test_loss:2.709, Lr:9.20E-05
Epoch: 7, Train_acc:13.9%, Train_loss:2.732, Test_acc:13.6%, Test_loss:2.693, Lr:9.20E-05
Epoch: 8, Train_acc:14.6%, Train_loss:2.723, Test_acc:13.6%, Test_loss:2.666, Lr:8.46E-05
Epoch: 9, Train_acc:14.9%, Train_loss:2.693, Test_acc:13.9%, Test_loss:2.660, Lr:8.46E-05
Epoch:10, Train_acc:15.4%, Train_loss:2.688, Test_acc:13.6%, Test_loss:2.652, Lr:8.46E-05
Epoch:11, Train_acc:15.4%, Train_loss:2.670, Test_acc:13.6%, Test_loss:2.634, Lr:8.46E-05
Epoch:12, Train_acc:15.8%, Train_loss:2.665, Test_acc:15.0%, Test_loss:2.618, Lr:7.79E-05
Epoch:13, Train_acc:16.1%, Train_loss:2.643, Test_acc:13.3%, Test_loss:2.611, Lr:7.79E-05
Epoch:14, Train_acc:16.9%, Train_loss:2.615, Test_acc:13.6%, Test_loss:2.605, Lr:7.79E-05
Epoch:15, Train_acc:15.8%, Train_loss:2.625, Test_acc:14.2%, Test_loss:2.587, Lr:7.79E-05
Epoch:16, Train_acc:17.1%, Train_loss:2.602, Test_acc:13.9%, Test_loss:2.569, Lr:7.16E-05
Epoch:17, Train_acc:16.0%, Train_loss:2.605, Test_acc:14.7%, Test_loss:2.560, Lr:7.16E-05
Epoch:18, Train_acc:17.3%, Train_loss:2.600, Test_acc:14.4%, Test_loss:2.556, Lr:7.16E-05
Epoch:19, Train_acc:17.8%, Train_loss:2.586, Test_acc:14.4%, Test_loss:2.541, Lr:7.16E-05
Epoch:20, Train_acc:18.1%, Train_loss:2.547, Test_acc:14.7%, Test_loss:2.551, Lr:6.59E-05
Epoch:21, Train_acc:17.5%, Train_loss:2.551, Test_acc:14.2%, Test_loss:2.533, Lr:6.59E-05
Epoch:22, Train_acc:18.3%, Train_loss:2.547, Test_acc:15.0%, Test_loss:2.535, Lr:6.59E-05
Epoch:23, Train_acc:18.7%, Train_loss:2.530, Test_acc:15.6%, Test_loss:2.536, Lr:6.59E-05
Epoch:24, Train_acc:17.4%, Train_loss:2.510, Test_acc:15.6%, Test_loss:2.509, Lr:6.06E-05
Epoch:25, Train_acc:18.3%, Train_loss:2.521, Test_acc:15.6%, Test_loss:2.508, Lr:6.06E-05
Epoch:26, Train_acc:16.8%, Train_loss:2.530, Test_acc:15.0%, Test_loss:2.497, Lr:6.06E-05
Epoch:27, Train_acc:19.0%, Train_loss:2.496, Test_acc:15.8%, Test_loss:2.502, Lr:6.06E-05
Epoch:28, Train_acc:18.6%, Train_loss:2.491, Test_acc:16.1%, Test_loss:2.470, Lr:5.58E-05
Epoch:29, Train_acc:18.8%, Train_loss:2.509, Test_acc:16.1%, Test_loss:2.497, Lr:5.58E-05
Epoch:30, Train_acc:19.3%, Train_loss:2.481, Test_acc:15.3%, Test_loss:2.474, Lr:5.58E-05
Epoch:31, Train_acc:19.7%, Train_loss:2.481, Test_acc:16.1%, Test_loss:2.472, Lr:5.58E-05
Epoch:32, Train_acc:20.0%, Train_loss:2.467, Test_acc:16.1%, Test_loss:2.480, Lr:5.13E-05
Epoch:33, Train_acc:19.7%, Train_loss:2.471, Test_acc:16.1%, Test_loss:2.465, Lr:5.13E-05
Epoch:34, Train_acc:19.4%, Train_loss:2.462, Test_acc:16.7%, Test_loss:2.460, Lr:5.13E-05
Epoch:35, Train_acc:18.3%, Train_loss:2.472, Test_acc:15.6%, Test_loss:2.465, Lr:5.13E-05
Epoch:36, Train_acc:21.0%, Train_loss:2.438, Test_acc:16.9%, Test_loss:2.454, Lr:4.72E-05
Epoch:37, Train_acc:20.8%, Train_loss:2.455, Test_acc:16.9%, Test_loss:2.479, Lr:4.72E-05
Epoch:38, Train_acc:18.1%, Train_loss:2.454, Test_acc:16.9%, Test_loss:2.471, Lr:4.72E-05
Epoch:39, Train_acc:20.6%, Train_loss:2.461, Test_acc:16.9%, Test_loss:2.476, Lr:4.72E-05
Epoch:40, Train_acc:20.3%, Train_loss:2.446, Test_acc:17.2%, Test_loss:2.458, Lr:4.34E-05

四、结果可视化

1. Loss与Accuracy图

#--------------------------------结果可视化------------------------
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)

if len(train_acc) > 0 and len(test_acc) > 0:
    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()
else:
    print("No data available for plotting.")

代码输出：

2. 指定图片进行预测

输入代码：

from PIL import Image 

classes = list(total_data.class_to_idx)

def predict_one_image(image_path, model, transform, classes):
    
    test_img = Image.open(image_path).convert('RGB')
    plt.imshow(test_img)  # 展示预测的图片

    test_img = transform(test_img)
    img = test_img.to(device).unsqueeze(0)
    
    model.eval()
    output = model(img)

    _,pred = torch.max(output,1)
    pred_class = classes[pred]
    print(f'预测结果是：{pred_class}')
    # 预测训练集中的某张照片
predict_one_image(image_path='./6-data/Angelina Jolie/001_fe3347c0.jpg', 
                  model=model, 
                  transform=train_transforms, 
                  classes=classes)

输出代码：

预测结果是：Angelina Jolie

五、总结

本周练习过程中，运行代码有些不顺利，但一件件的解决了。
首先，学习了如何设置动态学习率。
其次也学习了关于如何调用VGG-16模型
最后，学习了相对于上周增加指定图片预测与保存并加载模型这个两个模块。