第P6周：VGG-16算法-Pytorch实现人脸识别

原创已于 2025-07-25 22:40:38 修改 · 780 阅读

9 ·

CC 4.0 BY-SA版权

文章标签：

#pytorch #人工智能 #python

于 2025-07-25 22:35:26 首次发布

AI学习专栏收录该内容

6 篇文章

订阅专栏

部署运行你感兴趣的模型镜像

🍨 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
我的环境：

语言环境：Python3.12
编译器：PyCharm
深度学习环境：
torch==1.12.1+cu113
torchvision==0.13.1+cu113

一、实验

1、目的

保存训练过程中的最佳模型权重
调用官方的VGG-16网络框架

2、总结

1）掌握VGG-16网络基本使用

在训练过程中，VGG16一般采用基于随机梯度下降（Stochastic Gradient Descent，SGD）的反向传播算法，通过最小化交叉熵损失函数来优化模型参数。在训练过程中，可以使用数据增强、正则化、dropout等技术来提高模型的泛化能力和鲁棒性。总的来说，VGG16是一个非常经典和有效的卷积神经网络模型，具有良好的特征提取和分类能力，可以应用于图像分类、目标检测等计算机视觉任务。

2）层划分

输入：224x224的RGB 彩色图像；
block1：包含2个 [64x3x3] 的卷积层；
block2：包含2个 [128x3x3] 的卷积层；
block3：包含3个 [256x3x3] 的卷积层；
block4：包含3个 [512x3x3] 的卷积层；
block5：包含3个 [512x3x3] 的卷积层；
接着有3个全连接层；
一个分类输出层，经过 SoftMax 输出 1000个类的后验概率。

3）参数展开过程图解

上图中神经网络划分了5个block； - 红色：下采样（Max Pooling）； - 白色：卷积层+ReLU； - 蓝色：全连接+ReLU； - 神经网络参数传递从左向右； - 参数传递过程中，通道数越来越深，尺寸越来越小； - 从7x7x512下采样后，参数被拉平为1x1的长向量，进入全连接层；

4.）VGG 16各层参数数量

从图中可以看出，第一个全连接nfcr参数数量最多，达到了1亿多，这是因为前一个卷积层输出的向量在拉长成25088维向量，第一个全连接层4096要与25088相乘（每个神经元都要相连）。

5）个人理解

VGG16模型的设计思想是通过堆叠多个较小的卷积层和池化层来构建深层网络，以增强模型的表达能力。具体来说，VGG16模型由16个卷积层和3个全连接层组成。其中，卷积层主要用于提取输入图像的特征，而全连接层则用于将提取到的特征映射到类别概率上。 VGG16的卷积部分采用了较小的3x3卷积核和步长为1的卷积操作，这种设计方式使得网络可以更深，从而提升了特征的表达能力。

3、结果

D:\Programs\Python\Python39\python.exe D:\PycharmProjects\PythonProject\P5\main.py 
cpu
['test', 'train']
{'adidas': 0, 'nike': 1}
Shape of X [N, C, H, W]:  torch.Size([32, 3, 224, 224])
Shape of y:  torch.Size([32]) torch.int64
Using cpu device
Model(
  (conv1): Sequential(
    (0): Conv2d(3, 12, kernel_size=(5, 5), stride=(1, 1))
    (1): BatchNorm2d(12, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
  )
  (conv2): Sequential(
    (0): Conv2d(12, 12, kernel_size=(5, 5), stride=(1, 1))
    (1): BatchNorm2d(12, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
  )
  (pool3): Sequential(
    (0): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (conv4): Sequential(
    (0): Conv2d(12, 24, kernel_size=(5, 5), stride=(1, 1))
    (1): BatchNorm2d(24, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
  )
  (conv5): Sequential(
    (0): Conv2d(24, 24, kernel_size=(5, 5), stride=(1, 1))
    (1): BatchNorm2d(24, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
  )
  (pool6): Sequential(
    (0): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (dropout): Sequential(
    (0): Dropout(p=0.2, inplace=False)
  )
  (fc): Sequential(
    (0): Linear(in_features=60000, out_features=2, bias=True)
  )
)
Epoch: 1, Train_acc:52.4%, Train_loss:0.725, Test_acc:64.5%, Test_loss:0.690, Lr:1.00E-04
Epoch: 2, Train_acc:62.5%, Train_loss:0.663, Test_acc:64.5%, Test_loss:0.657, Lr:9.20E-05
Epoch: 3, Train_acc:63.5%, Train_loss:0.635, Test_acc:61.8%, Test_loss:0.731, Lr:9.20E-05
Epoch: 4, Train_acc:65.5%, Train_loss:0.600, Test_acc:64.5%, Test_loss:0.603, Lr:8.46E-05
Epoch: 5, Train_acc:73.3%, Train_loss:0.557, Test_acc:64.5%, Test_loss:0.613, Lr:8.46E-05
Epoch: 6, Train_acc:73.7%, Train_loss:0.532, Test_acc:60.5%, Test_loss:0.685, Lr:7.79E-05
Epoch: 7, Train_acc:74.9%, Train_loss:0.523, Test_acc:69.7%, Test_loss:0.566, Lr:7.79E-05
Epoch: 8, Train_acc:77.3%, Train_loss:0.503, Test_acc:72.4%, Test_loss:0.573, Lr:7.16E-05
Epoch: 9, Train_acc:76.7%, Train_loss:0.500, Test_acc:69.7%, Test_loss:0.552, Lr:7.16E-05
Epoch:10, Train_acc:81.1%, Train_loss:0.469, Test_acc:67.1%, Test_loss:0.561, Lr:6.59E-05
Epoch:11, Train_acc:82.1%, Train_loss:0.453, Test_acc:71.1%, Test_loss:0.508, Lr:6.59E-05
Epoch:12, Train_acc:81.5%, Train_loss:0.449, Test_acc:69.7%, Test_loss:0.515, Lr:6.06E-05
Epoch:13, Train_acc:84.7%, Train_loss:0.426, Test_acc:72.4%, Test_loss:0.536, Lr:6.06E-05
Epoch:14, Train_acc:85.9%, Train_loss:0.412, Test_acc:69.7%, Test_loss:0.565, Lr:5.58E-05
Epoch:15, Train_acc:85.5%, Train_loss:0.409, Test_acc:72.4%, Test_loss:0.586, Lr:5.58E-05
Epoch:16, Train_acc:86.5%, Train_loss:0.389, Test_acc:71.1%, Test_loss:0.534, Lr:5.13E-05
Epoch:17, Train_acc:86.7%, Train_loss:0.389, Test_acc:73.7%, Test_loss:0.535, Lr:5.13E-05
Epoch:18, Train_acc:88.8%, Train_loss:0.374, Test_acc:69.7%, Test_loss:0.493, Lr:4.72E-05
Epoch:19, Train_acc:87.8%, Train_loss:0.382, Test_acc:72.4%, Test_loss:0.528, Lr:4.72E-05
Epoch:20, Train_acc:88.4%, Train_loss:0.364, Test_acc:75.0%, Test_loss:0.504, Lr:4.34E-05
Epoch:21, Train_acc:90.4%, Train_loss:0.360, Test_acc:71.1%, Test_loss:0.492, Lr:4.34E-05
Epoch:22, Train_acc:88.6%, Train_loss:0.362, Test_acc:72.4%, Test_loss:0.481, Lr:4.00E-05
Epoch:23, Train_acc:91.6%, Train_loss:0.347, Test_acc:75.0%, Test_loss:0.486, Lr:4.00E-05
Epoch:24, Train_acc:90.0%, Train_loss:0.348, Test_acc:75.0%, Test_loss:0.489, Lr:3.68E-05
Epoch:25, Train_acc:89.2%, Train_loss:0.351, Test_acc:76.3%, Test_loss:0.498, Lr:3.68E-05
Epoch:26, Train_acc:91.6%, Train_loss:0.339, Test_acc:75.0%, Test_loss:0.501, Lr:3.38E-05
Epoch:27, Train_acc:90.6%, Train_loss:0.338, Test_acc:75.0%, Test_loss:0.494, Lr:3.38E-05
Epoch:28, Train_acc:91.8%, Train_loss:0.329, Test_acc:76.3%, Test_loss:0.483, Lr:3.11E-05
Epoch:29, Train_acc:91.0%, Train_loss:0.333, Test_acc:76.3%, Test_loss:0.528, Lr:3.11E-05
Epoch:30, Train_acc:91.8%, Train_loss:0.327, Test_acc:76.3%, Test_loss:0.486, Lr:2.86E-05
Epoch:31, Train_acc:92.2%, Train_loss:0.318, Test_acc:76.3%, Test_loss:0.493, Lr:2.86E-05
Epoch:32, Train_acc:91.8%, Train_loss:0.322, Test_acc:77.6%, Test_loss:0.514, Lr:2.63E-05
Epoch:33, Train_acc:91.0%, Train_loss:0.321, Test_acc:77.6%, Test_loss:0.498, Lr:2.63E-05
Epoch:34, Train_acc:92.4%, Train_loss:0.305, Test_acc:77.6%, Test_loss:0.491, Lr:2.42E-05
Epoch:35, Train_acc:92.8%, Train_loss:0.317, Test_acc:76.3%, Test_loss:0.568, Lr:2.42E-05
Epoch:36, Train_acc:91.8%, Train_loss:0.312, Test_acc:76.3%, Test_loss:0.482, Lr:2.23E-05
Epoch:37, Train_acc:93.6%, Train_loss:0.304, Test_acc:76.3%, Test_loss:0.441, Lr:2.23E-05
Epoch:38, Train_acc:92.4%, Train_loss:0.301, Test_acc:76.3%, Test_loss:0.480, Lr:2.05E-05
Epoch:39, Train_acc:93.4%, Train_loss:0.304, Test_acc:76.3%, Test_loss:0.485, Lr:2.05E-05
Epoch:40, Train_acc:93.0%, Train_loss:0.299, Test_acc:77.6%, Test_loss:0.456, Lr:1.89E-05
Done
预测结果是：adidas

进程已结束，退出代码为 0

二、源代码


#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,warnings

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

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

#2. 导入数据
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]
print(classeNames)

# 关于transforms.Compose的更多介绍可以参考：https://blog.youkuaiyun.com/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。
])

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

#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])
print(train_dataset)
print(test_dataset)
print(len(train_dataset))
print(len(test_dataset))

batch_size = 32

train_dl = torch.utils.data.DataLoader(train_dataset,
                                           batch_size=batch_size,
                                           shuffle=True,
                                           num_workers=0)    #   num_workers=1 会报错
test_dl = torch.utils.data.DataLoader(test_dataset,
                                          batch_size=batch_size,
                                          shuffle=True,
                                          num_workers=0)    #   num_workers=1 会报错

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

#二、调用官方的VGG-16模型

# VGG-16结构说明：
#
# ● 13个卷积层（Convolutional Layer），分别用blockX_convX表示；
# ● 3个全连接层（Fully connected Layer），用classifier表示；
# ● 5个池化层（Pool layer）

from torchvision.models import vgg16

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(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)
print(model)


#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

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 0.98
#     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)

#调用官方动态学习率接口

# 调用官方动态学习率接口时使用
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

import copy

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

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):
    # 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)

    # 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(model.state_dict(), PATH)

print('Done')


#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, 4))
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()


#2. 指定图片进行预测
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='./data/Angelina Jolie/001_fe3347c0.jpg',
                  model=model,
                  transform=train_transforms,
                  classes=classes)

#3. 模型评估

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

print(epoch_test_acc)
print(epoch_test_loss)

您可能感兴趣的与本文相关的镜像

PyTorch 2.5

PyTorch

Cuda

PyTorch 是一个开源的 Python 机器学习库，基于 Torch 库，底层由 C++ 实现，应用于人工智能领域，如计算机视觉和自然语言处理