- 🍨 本文为🔗365天深度学习训练营 中的学习记录博客
- 🍦 参考文章:Pytorch实战 | 第P3周:彩色图片识别:天气识别
- 🍖 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
一、前期准备
1.设置GPU
import torch # 导入PyTorch库
import torch.nn as nn # 导入PyTorch神经网络模块,包含了为建立神经网络模型而预定义好的各种层、损失函数等。
import matplotlib.pyplot as plt # 用于绘制图像和可视化
from torchvision import transforms, datasets # 导入图像变换和数据集
import os, PIL, pathlib # 导入操作系统、图像处理和路径处理库
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') # 检查是否有GPU支持,并设置为默认设备
print(device) # 打印当前使用的设备('cuda' or 'cpu')
执行结果:
cuda
2. 导入数据
data_dir = pathlib.Path('./data/') # 使用pathlib.Path处理路径,它提供了更多关于路径的方法和属性
data_paths = list(data_dir.glob('*')) # 使用Path.glob方法搜索当前目录下的所有文件/文件夹。这个方法返回一个迭代器,所以使用list()将其转换为列表。
classNames = [path.name for path in data_paths] # 使用列表推导式从每个路径中提取名称(即最后一个部分)。例如,路径./data/cloudy将提取为cloudy。
print(classNames)
执行结果:
['cloudy', 'rain', 'shine', 'sunrise']
3.预处理和加载数据集
total_datadir = './data/'
## train_transforms 进行图像预处理
train_transforms = transforms.Compose([
transforms.Resize([224, 224]), # 将输入图片resize成统一尺寸224 x 224大小
transforms.ToTensor(), # 转换为tensor格式, 且归一化
transforms.Normalize( # 标准化处理-->转换为标准正态分布(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(total_datadir,transform=train_transforms) # 加载数据集并应用预处理
print(total_data)
执行结果:
Dataset ImageFolder
Number of datapoints: 1125
Root location: ./data/
StandardTransform
Transform: Compose(
Resize(size=[224, 224], interpolation=bilinear, max_size=None, antialias=warn)
ToTensor()
Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
)
4.划分数据集
train_size = int(0.8 * len(total_data)) # 训练数据集80%
test_size = len(total_data) - train_size # 测试集 20%
# 使用 torch.utils.data.random_split 将数据集分为训练集和测试集
train_dataset, test_dataset = torch.utils.data.random_split(total_data, [train_size, test_size])
print(train_dataset)
print(test_dataset)
print(train_size)
print(test_size)
执行结果:
<torch.utils.data.dataset.Subset object at 0x0000022A300C1220>
<torch.utils.data.dataset.Subset object at 0x0000022A3001D9A0>
900
225
5.数据加载器
batch_size = 32 # 定义每个批次中包含的样本数量
# 使用DataLoader创建训练和测试的数据加载器,用于从数据集中加载批次的数据
train_dl = torch.utils.data.DataLoader(train_dataset,
batch_size=batch_size,
shuffle=True,
num_workers=0)
test_dl = torch.utils.data.DataLoader(test_dataset,
batch_size=batch_size,
shuffle=True,
num_workers=0)
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
执行结果:
Shape of X [N, C, H, W]: torch.Size([32, 3, 224, 224])
Shape of y: torch.Size([32]) torch.int64
二、构建简单的CNN网络
定义一个带有批量归一化(Batch Normalization) 的卷积神经网络结构,该网络可以对输入的图像进行特征提取和分类。
卷积层:
**nn.Conv2d():**定义一个二维卷积层。
指定输入通道数(in_channels)、输出通道数(out_channels)、卷积核大小(kernel_size)、步长(stride)、填充(padding)。
批量归一化层:
**
nn.BatchNorm2d()
:**定义一个二维批量归一化层。用于规范化卷积层输出的分布,有助于加速收敛和提高模型性能。
最大池化层:
**
nn.MaxPool2d()
:**定义一个二维最大池化层。用于降低特征图的维度,保留重要的特征。
全连接层:
**
nn.Linear()
:**定义一个全连接(线性)层。用于将池化层输出映射到分类标签。
# 导入PyTorch的函数式模块,其中包含了一些常用的非线性激活函数,如ReLU。
import torch.nn.functional as F
# 定义网络模型:
class Network_bn(nn.Module): # 定义了一个名为'Network_bn'的类,继承自'nn.Module',这是Pytorch中构建神经网络模型的基类。
def __init__(self): # 构造函数,用于初始化网络结构和参数
super(Network_bn, self).__init__() # 调用父类(nn.Module)的构造函数
self.conv1 = nn.Conv2d(in_channels=3, out_channels=12, kernel_size=5, stride=1, padding=0)
self.bn1 = nn.BatchNorm2d(12)
self.conv2 = nn.Conv2d(in_channels=12, out_channels=12, kernel_size=5, stride=1, padding=0)
self.bn2 = nn.BatchNorm2d(12)
self.pool = nn.MaxPool2d(2,2)
self.conv4 = nn.Conv2d(in_channels=12, out_channels=24, kernel_size=5, stride=1, padding=0)
self.bn4 = nn.BatchNorm2d(24)
self.conv5 = nn.Conv2d(in_channels=24, out_channels=24, kernel_size=5, stride=1, padding=0)
self.bn5 = nn.BatchNorm2d(24)
self.fc1 = nn.Linear(24*50*50, len(classNames))
def forward(self, x): # 前向传播函数
x = F.relu(self.bn1(self.conv1(x))) # 对输入应用卷积、批量归一化和ReLU激活函数。
x = F.relu(self.bn2(self.conv2(x)))
x = self.pool(x) # 使用最大池化层对特征图进行下采样。在网络的不同部分进行,形成特征提取过程。
x = F.relu(self.bn4(self.conv4(x)))
x = F.relu(self.bn5(self.conv5(x)))
x = self.pool(x)
x = x.view(-1, 24*50*50)
x = self.fc1(x) # 最后进行全连接层,将提取的特征映射到分类标签。
return x
device = "cuda" if torch.cuda.is_available() else "cpu"
print("Using {} device".format(device))
model = Network_bn().to(device) # 创建Network_bn的实例,将模型移动到指定的设备上。
print(model)
执行结果:
Using cuda device
Network_bn(
(conv1): Conv2d(3, 12, kernel_size=(5, 5), stride=(1, 1))
(bn1): BatchNorm2d(12, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv2): Conv2d(12, 12, kernel_size=(5, 5), stride=(1, 1))
(bn2): BatchNorm2d(12, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(pool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(conv4): Conv2d(12, 24, kernel_size=(5, 5), stride=(1, 1))
(bn4): BatchNorm2d(24, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv5): Conv2d(24, 24, kernel_size=(5, 5), stride=(1, 1))
(bn5): BatchNorm2d(24, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(fc1): Linear(in_features=60000, out_features=4, bias=True)
)
三、训练模型
1.设置超参数
loss_fn = nn.CrossEntropyLoss() # 创建交叉熵损失函数,用于计算模型输出与真实标签之间的差距
learn_rate = 1e-4 # 学习率,即优化器在更新模型参数时使用的步长
opt = torch.optim.SGD(model.parameters(), lr=learn_rate) # opt, 创建SGD优化器,用于更新模型的参数。(传入模型的参数和学习率)
2.编写训练函数
# 训练循环
def train(dataloader, model, loss_fn, optimizer):
size = len(dataloader.dataset) # 获取训练集的大小
num_batches = len(dataloader) # 获取批次数量
train_loss, train_acc = 0, 0 # 初始化训练损失和正确率
for X, y in dataloader: # 遍历训练集的每个批次,获取图片及其标签
X, y = X.to(device), y.to(device) # 将数据移动到设备上
# 计算预测误差
pred = model(X) # 网络输出,前向传播得到模型预测值
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,清除之前的梯度
loss.backward() # 反向传播,计算梯度
optimizer.step() # 每一步自动更新模型参数
# 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
3.编写测试函数
def test (dataloader, model, loss_fn):
size = len(dataloader.dataset) # 获取测试集的大小
num_bathes = len(dataloader) # 获取批次数量
test_loss, test_acc = 0, 0 # 初始化测试损失和测试准确率
# 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(): # 在测试过程中,不需要计算梯度
for imgs, target in dataloader: 遍历测试数据集的每个批次
imgs, target = imgs.to(device), target.to(device) # 将数据移动到设备上
# 计算loss
target_pred = model(imgs) # 模型预测
loss = loss_fn(target_pred, target) # 计算损失
test_loss += loss.item()
test_acc += (target_pred.argmax(1) == target).type(torch.float).sum().item()
test_acc /= size
test_loss /= num_bathes
return test_acc, test_loss
4.正式训练
epochs = 20
train_loss = []
train_acc = []
test_loss = []
test_acc = []
for epoch in range(epochs):
model.train() # 设置模型为训练模式
epoch_train_acc, epoch_train_loss = train(train_dl, model, loss_fn, opt)
model.eval() # 设置模型为评估模式,不启用Batch Normalization 和Dropout
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)
template = ('Epoch:{:2d}, Train_acc:{:.1f}%, Train_loss:{:.3f}, Test_acc:{:.1f}%, Test_loss:{:.3f}')
print(template.format(epoch+1, epoch_train_acc*100, epoch_train_loss, epoch_test_acc*100, epoch_test_loss))
print("Done")
执行结果:
Epoch: 1, Train_acc:61.4%, Train_loss:0.983, Test_acc:61.8%, Test_loss:1.203
Epoch: 2, Train_acc:81.2%, Train_loss:0.637, Test_acc:80.0%, Test_loss:0.582
Epoch: 3, Train_acc:85.3%, Train_loss:0.529, Test_acc:78.2%, Test_loss:0.660
Epoch: 4, Train_acc:84.1%, Train_loss:0.473, Test_acc:86.7%, Test_loss:0.444
Epoch: 5, Train_acc:88.1%, Train_loss:0.400, Test_acc:77.3%, Test_loss:0.499
Epoch: 6, Train_acc:88.9%, Train_loss:0.376, Test_acc:88.9%, Test_loss:0.440
Epoch: 7, Train_acc:90.7%, Train_loss:0.333, Test_acc:89.8%, Test_loss:0.330
Epoch: 8, Train_acc:90.2%, Train_loss:0.310, Test_acc:90.2%, Test_loss:0.610
Epoch: 9, Train_acc:91.7%, Train_loss:0.292, Test_acc:91.6%, Test_loss:0.307
Epoch:10, Train_acc:91.6%, Train_loss:0.273, Test_acc:89.8%, Test_loss:0.317
Epoch:11, Train_acc:91.4%, Train_loss:0.260, Test_acc:90.2%, Test_loss:0.466
Epoch:12, Train_acc:92.2%, Train_loss:0.253, Test_acc:92.0%, Test_loss:0.274
Epoch:13, Train_acc:92.6%, Train_loss:0.238, Test_acc:91.6%, Test_loss:0.273
Epoch:14, Train_acc:93.6%, Train_loss:0.241, Test_acc:91.6%, Test_loss:0.260
Epoch:15, Train_acc:93.9%, Train_loss:0.219, Test_acc:92.4%, Test_loss:0.283
Epoch:16, Train_acc:94.7%, Train_loss:0.197, Test_acc:91.1%, Test_loss:0.268
Epoch:17, Train_acc:94.6%, Train_loss:0.209, Test_acc:92.4%, Test_loss:0.249
Epoch:18, Train_acc:94.4%, Train_loss:0.198, Test_acc:92.4%, Test_loss:0.264
Epoch:19, Train_acc:93.7%, Train_loss:0.206, Test_acc:92.0%, Test_loss:0.267
Epoch:20, Train_acc:95.7%, Train_loss:0.173, Test_acc:92.4%, Test_loss:0.246
Done
四、结果可视化
# 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()