第P6周：VGG-16算法实现人脸识别

• 🍨 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 | 接辅导、项目定制
• 🚀 文章来源：K同学的学习圈子

目录

代码及运行结果：

一、 前期准备

二、调用官方的VGG-16模型

 

 三、 训练模型

四、 结果可视化

 个人总结：

---

代码及运行结果：

一、 前期准备

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)

import os,PIL,random,pathlib

data_dir = './6-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("./6-data/",transform=train_transforms)
print(total_data)

print(total_data.class_to_idx)

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])

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)

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

cuda
['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']
Dataset ImageFolder
    Number of datapoints: 1801
    Root location: ./6-data/
    StandardTransform
Transform: Compose(
               Resize(size=[224, 224], interpolation=bilinear)
               ToTensor()
               Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
           )
{'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}
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

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 = nn.Sequential(
    nn.Linear(512 * 7 * 7, 1024),
    nn.ReLU(inplace=True),
    nn.Dropout(0.5,inplace=False),
    nn.Linear(1024,512),
    nn.ReLU(inplace=True),
    nn.Dropout(0.5),
    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) 
print(model)

 

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=1024, bias=True)
    (1): ReLU(inplace=True)
    (2): Dropout(p=0.5, inplace=False)
    (3): Linear(in_features=1024, out_features=512, bias=True)
    (4): ReLU(inplace=True)
    (5): Dropout(p=0.5, inplace=False)
    (6): Linear(in_features=512, out_features=17, bias=True)
  )
)

 三、 训练模型

# 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

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

learn_rate = 1e-3 # 初始学习率
# 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

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')

 

Epoch: 1, Train_acc:6.9%, Train_loss:2.824, Test_acc:12.7%, Test_loss:2.784, Lr:1.00E-03
Epoch: 2, Train_acc:13.6%, Train_loss:2.767, Test_acc:16.3%, Test_loss:2.731, Lr:1.00E-03
Epoch: 3, Train_acc:16.2%, Train_loss:2.707, Test_acc:17.2%, Test_loss:2.685, Lr:1.00E-03
Epoch: 4, Train_acc:16.6%, Train_loss:2.650, Test_acc:18.3%, Test_loss:2.631, Lr:9.20E-04
Epoch: 5, Train_acc:18.6%, Train_loss:2.588, Test_acc:18.3%, Test_loss:2.574, Lr:9.20E-04
Epoch: 6, Train_acc:20.4%, Train_loss:2.523, Test_acc:19.9%, Test_loss:2.523, Lr:9.20E-04
Epoch: 7, Train_acc:19.4%, Train_loss:2.479, Test_acc:19.4%, Test_loss:2.450, Lr:9.20E-04
Epoch: 8, Train_acc:22.4%, Train_loss:2.421, Test_acc:21.1%, Test_loss:2.427, Lr:8.46E-04
Epoch: 9, Train_acc:23.4%, Train_loss:2.364, Test_acc:22.7%, Test_loss:2.362, Lr:8.46E-04
Epoch:10, Train_acc:26.7%, Train_loss:2.306, Test_acc:24.4%, Test_loss:2.343, Lr:8.46E-04
Epoch:11, Train_acc:27.2%, Train_loss:2.272, Test_acc:26.3%, Test_loss:2.284, Lr:8.46E-04
Epoch:12, Train_acc:28.5%, Train_loss:2.241, Test_acc:26.3%, Test_loss:2.234, Lr:7.79E-04
Epoch:13, Train_acc:28.5%, Train_loss:2.187, Test_acc:27.4%, Test_loss:2.212, Lr:7.79E-04
Epoch:14, Train_acc:29.1%, Train_loss:2.172, Test_acc:28.8%, Test_loss:2.187, Lr:7.79E-04
Epoch:15, Train_acc:31.6%, Train_loss:2.116, Test_acc:30.2%, Test_loss:2.160, Lr:7.79E-04
Epoch:16, Train_acc:32.1%, Train_loss:2.089, Test_acc:31.9%, Test_loss:2.136, Lr:7.16E-04
Epoch:17, Train_acc:34.2%, Train_loss:2.075, Test_acc:32.7%, Test_loss:2.102, Lr:7.16E-04
Epoch:18, Train_acc:36.2%, Train_loss:2.016, Test_acc:34.6%, Test_loss:2.073, Lr:7.16E-04
Epoch:19, Train_acc:34.3%, Train_loss:2.013, Test_acc:36.6%, Test_loss:2.064, Lr:7.16E-04
Epoch:20, Train_acc:36.0%, Train_loss:1.974, Test_acc:38.2%, Test_loss:2.031, Lr:6.59E-04
Epoch:21, Train_acc:37.0%, Train_loss:1.950, Test_acc:39.1%, Test_loss:2.007, Lr:6.59E-04
Epoch:22, Train_acc:36.9%, Train_loss:1.932, Test_acc:40.2%, Test_loss:2.004, Lr:6.59E-04
Epoch:23, Train_acc:39.9%, Train_loss:1.901, Test_acc:40.2%, Test_loss:1.982, Lr:6.59E-04
Epoch:24, Train_acc:40.5%, Train_loss:1.889, Test_acc:41.3%, Test_loss:1.952, Lr:6.06E-04
Epoch:25, Train_acc:42.2%, Train_loss:1.859, Test_acc:40.4%, Test_loss:1.939, Lr:6.06E-04
Epoch:26, Train_acc:43.3%, Train_loss:1.842, Test_acc:39.3%, Test_loss:1.911, Lr:6.06E-04
Epoch:27, Train_acc:44.2%, Train_loss:1.803, Test_acc:40.7%, Test_loss:1.897, Lr:6.06E-04
Epoch:28, Train_acc:43.8%, Train_loss:1.794, Test_acc:41.6%, Test_loss:1.885, Lr:5.58E-04
Epoch:29, Train_acc:44.4%, Train_loss:1.791, Test_acc:42.9%, Test_loss:1.853, Lr:5.58E-04
Epoch:30, Train_acc:45.9%, Train_loss:1.759, Test_acc:43.2%, Test_loss:1.859, Lr:5.58E-04
Epoch:31, Train_acc:46.5%, Train_loss:1.738, Test_acc:43.5%, Test_loss:1.844, Lr:5.58E-04
Epoch:32, Train_acc:47.4%, Train_loss:1.721, Test_acc:44.0%, Test_loss:1.808, Lr:5.13E-04
Epoch:33, Train_acc:48.6%, Train_loss:1.698, Test_acc:45.4%, Test_loss:1.827, Lr:5.13E-04
Epoch:34, Train_acc:49.3%, Train_loss:1.672, Test_acc:44.9%, Test_loss:1.802, Lr:5.13E-04
Epoch:35, Train_acc:50.0%, Train_loss:1.652, Test_acc:45.7%, Test_loss:1.792, Lr:5.13E-04
Epoch:36, Train_acc:49.9%, Train_loss:1.637, Test_acc:45.4%, Test_loss:1.765, Lr:4.72E-04
Epoch:37, Train_acc:50.6%, Train_loss:1.640, Test_acc:46.8%, Test_loss:1.772, Lr:4.72E-04
Epoch:38, Train_acc:51.3%, Train_loss:1.613, Test_acc:47.4%, Test_loss:1.763, Lr:4.72E-04
Epoch:39, Train_acc:51.7%, Train_loss:1.610, Test_acc:46.3%, Test_loss:1.740, Lr:4.72E-04
Epoch:40, Train_acc:52.6%, Train_loss:1.583, Test_acc:47.1%, Test_loss:1.740, Lr:4.34E-04
Done

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()

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)

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

print('epoch_test_acc, epoch_test_loss:',epoch_test_acc, epoch_test_loss)

# 查看是否与我们记录的最高准确率一致
epoch_test_acc

预测结果是：Angelina Jolie
epoch_test_acc, epoch_test_loss: 0.47368421052631576 1.754202703634898

0.47368421052631576

 

 个人总结：

        这次任务我主要将时间花费在如何提高模型准确率上，参考文档中给出的代码在测试集的准确率只有不到30%，于是我就参考训练营中往届大佬发出的文章，对代码就行了一下的优化：

• 修改VGG-16网络中的全连接部分。
  • 减少了全连接层参数
    • 期间使用金字塔原则设置了一次参数
    • 又试了别人给出的参数
  • 增加dropout层，减少过拟合
• 将初始学习率提高为1e-3，进一步提升模型的准确率
• 同时不断更改学习率的衰减速度，从0.92到0.98和0.88，尝试不同的参数

最后成功将准确率提高到50%以上。