- 🍨 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P4周：猴痘病识别
- 🍖 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
import torch.nn as nn
import matplotlib.pyplot as plt
from torchvision import transforms, datasets, models
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(device)
执行结果：
cuda
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('*'))
classNames = [str(path).split("\\")[1] for path in data_paths]
print(classNames)
执行结果：
['Monkeypox', 'Others']
3.图像预处理和加载数据集
total_datadir = './data/'
train_transforms = transforms.Compose([
transforms.Resize([224, 224]), # 将输入图片resize成统一尺寸224 x 224
transforms.ToTensor(), # 转换为tensor格式, 并归一化
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], # 用于图像标准化，RGB通道均值和标准差
std=[0.229, 0.224, 0.225])
])
# 从total_datadir指定的目录加载图像数据，并使用train_transforms定义的转换来处理这些图像。加载后的数据集total_data会包含处理后的图像及其对应的标签。
total_data = datasets.ImageFolder(total_datadir, transform=train_transforms)
print(total_data)
print(total_data.class_to_idx) # 打印出一个字典，显示整数标签与实际类名之间的映射关系
执行结果：
Dataset ImageFolder # 图像分类任务数据加载器
Number of datapoints: 2142 # 数据集包含2142张图片
Root location: ./data/
StandardTransform # 表示已对数据集应用了标准的转换
Transform: Compose( # 应用于数据集的转换序列
Resize(size=[224, 224], interpolation=bilinear, max_size=None, antialias=warn) # 使用双线性插值方法进行缩放
ToTensor()
Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
)
{'Monkeypox': 0, 'Others': 1}
4.划分数据集
train_size = int(0.8 * len(total_data))
test_size = len(total_data) - train_size
# torch.utils.data.random_split: 这是PyTorch提供的一个函数，用于将一个数据集随机分割为多个非重叠的新数据集。
train_dataset, test_dataset = torch.utils.data.random_split(total_data,[train_size, test_size])
print(train_dataset, test_dataset)
print(train_size, test_size)
执行结果：
成功从原始数据集中创建了两个子集，会显示这两个子集的内存地址。
<torch.utils.data.dataset.Subset object at 0x0000029E2CE57D60>
<torch.utils.data.dataset.Subset object at 0x0000029E2CE65970>
1713 429
5.数据加载器
在windows系统中需要将进程数设置为单进程，所以需要将数据处理部分DataLoader中的参数 num_workers设置为0或者采用默认为0的设置。
缺点：速度慢。
batch_size = 32
train_dl = torch.utils.data.DataLoader(train_dataset,
batch_size=batch_size,
shuffle=True)
test_dl = torch.utils.data.DataLoader(test_dataset,
batch_size=batch_size,
shuffle=True)
for imgs, labels in test_dl: # imgs：数据加载器中输入数据（图像）, # labels：数据的标签
# N：批次中图像数量
# C：图像通道数（RGB）
# H：图像高度
# W：图像宽度
print("Shape of imgs [N, C, H, W]: ", imgs.shape)
print("Shape of labels: ", labels.shape, labes.dtype) # 标签形状、标签数据类型
break # 只打印第一批
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
6.数据可视化【预警！！！！！！！！！！！！！！！！！】
plt.figure(figsize=(20,4))
for i in range(20):
imgs, labels = train_dataset[i]
plt.subplot(2, 10, i+1)
plt.imshow(imgs.permute(1,2,0))
plt.title(classNames[labels])
plt.axis('off')
plt.show()
执行结果：
二、构建简单的CNN网络
import torch.nn.functional as F
class Network_bn(nn.Module):
def __init__(self):
super(Network_bn, self).__init__()
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):
x = F.relu(self.bn1(self.conv1(x)))
x = F.relu(self.bn2(self.conv2(x)))
x = self.pool(x)
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)
return x
device = "cuda" if torch.cuda.is_available() else "cpu"
print("Using {} device".format(device))
model = Network_bn().to(device)
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=2, bias=True)
)
三、训练模型
1.设置超参数
loss_fn = nn.CrossEntropyLoss() # 创建损失函数
learn_rate = 1e-4 # 学习率
opt = torch.optim.SGD(model.parameters(), lr=learn_rate)
2.编写训练函数
def train(dataloader, model, loss_fn, optimizer):
size = len(dataloader.dataset) # 训练集大小
num_batches = len(dataloader) # 批次数目
train_loss, train_acc = 0, 0 # 初始化训练损失和正确率
for X, y in dataloader: # 获取图片标签
X, y = X.to(device), y.to(device)
# 计算预测误差
pred = model(X)
loss = loss_fn(pred, y)
# 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
3.编写测试函数
def test(dataloader, model, loss_fn):
size = len(dataloader.dataset) # 测试集大小
num_batches = len(dataloader) # 批次数目
test_loss, test_acc = 0, 0
# 当不进行训练是，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
4.正式训练
epochs = 20
train_loss = []
train_acc = []
test_loss = []
test_acc = []
for epoch in range(epochs):
model.train()
epoch_train_acc, epoch_train_loss = train(train_dl, model, loss_fn, opt)
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)
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:0.673, Test_acc:68.5%, Test_loss:0.623
Epoch: 2, Train_acc:68.6%, Train_loss:0.597, Test_acc:69.2%, Test_loss:0.592
Epoch: 3, Train_acc:72.6%, Train_loss:0.547, Test_acc:72.3%, Test_loss:0.561
Epoch: 4, Train_acc:76.9%, Train_loss:0.506, Test_acc:73.9%, Test_loss:0.556
Epoch: 5, Train_acc:78.9%, Train_loss:0.478, Test_acc:78.1%, Test_loss:0.514
Epoch: 6, Train_acc:81.0%, Train_loss:0.445, Test_acc:69.5%, Test_loss:0.617
Epoch: 7, Train_acc:80.9%, Train_loss:0.432, Test_acc:78.3%, Test_loss:0.504
Epoch: 8, Train_acc:83.2%, Train_loss:0.407, Test_acc:77.9%, Test_loss:0.507
Epoch: 9, Train_acc:86.0%, Train_loss:0.378, Test_acc:78.6%, Test_loss:0.442
Epoch:10, Train_acc:86.5%, Train_loss:0.367, Test_acc:79.5%, Test_loss:0.439
Epoch:11, Train_acc:86.9%, Train_loss:0.356, Test_acc:80.0%, Test_loss:0.439
Epoch:12, Train_acc:88.1%, Train_loss:0.344, Test_acc:82.8%, Test_loss:0.429
Epoch:13, Train_acc:88.3%, Train_loss:0.329, Test_acc:80.7%, Test_loss:0.405
Epoch:14, Train_acc:89.4%, Train_loss:0.319, Test_acc:83.9%, Test_loss:0.402
Epoch:15, Train_acc:89.7%, Train_loss:0.306, Test_acc:84.6%, Test_loss:0.395
Epoch:16, Train_acc:90.7%, Train_loss:0.297, Test_acc:83.7%, Test_loss:0.386
Epoch:17, Train_acc:91.2%, Train_loss:0.285, Test_acc:84.6%, Test_loss:0.386
Epoch:18, Train_acc:91.6%, Train_loss:0.280, Test_acc:84.4%, Test_loss:0.374
Epoch:19, Train_acc:92.2%, Train_loss:0.274, Test_acc:83.7%, Test_loss:0.369
Epoch:20, Train_acc:92.3%, Train_loss:0.268, Test_acc:83.2%, Test_loss:0.449
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
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()
执行结果：
2.指定图片进行预测
rom 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图片(Monkeypox)
predict_one_image(image_path='data/Monkeypox/M01_01_00.jpg',
model=model,
transform=train_transforms,
classes=classes)
# 预测训练集中的某张图片(Others)
predict_one_image(image_path='data/Others/NM01_01_00.jpg',
model=model,
transform=train_transforms,
classes=classes)
执行结果：
预测结果是：Monkeypox
预测结果是：Others
五、保存并加载模型
PATH = './model.pth' # 保存的参数文件名
torch.save(model.state_dict(), PATH)
# 将参数加载到model当中
model.load_state_dict(torch.load(PATH, map_location=device))
六、模型优化
尝试采用ResNet(Redidual Networks) :
class ResNetNetwork(nn.Module):
def __init__(self):
super(ResNetNetwork, self).__init__()
self.resnet = models.resnet18(pretrained=True)
num_ftrs = self.resnet.fc.in_features
self.resnet.fc = nn.Linear(num_ftrs, len(classNames))
def forward(self, x):
x = self.resnet(x)
return x
model = ResNetNetwork().to(device)
print(model)
修改学习率：
learn_rate = 1e-2 # 学习率
执行结果：
Epoch: 1, Train_acc:74.5%, Train_loss:0.491, Test_acc:86.2%, Test_loss:0.334
Epoch: 2, Train_acc:93.8%, Train_loss:0.191, Test_acc:92.8%, Test_loss:0.168
Epoch: 3, Train_acc:97.3%, Train_loss:0.099, Test_acc:95.8%, Test_loss:0.110
Epoch: 4, Train_acc:98.5%, Train_loss:0.062, Test_acc:95.6%, Test_loss:0.107
Epoch: 5, Train_acc:99.3%, Train_loss:0.035, Test_acc:95.3%, Test_loss:0.120
Epoch: 6, Train_acc:99.5%, Train_loss:0.027, Test_acc:96.5%, Test_loss:0.090
Epoch: 7, Train_acc:99.9%, Train_loss:0.016, Test_acc:96.5%, Test_loss:0.087
Epoch: 8, Train_acc:99.7%, Train_loss:0.016, Test_acc:95.6%, Test_loss:0.125
Epoch: 9, Train_acc:99.9%, Train_loss:0.009, Test_acc:96.7%, Test_loss:0.087
Epoch:10, Train_acc:100.0%, Train_loss:0.006, Test_acc:96.5%, Test_loss:0.077
Epoch:11, Train_acc:100.0%, Train_loss:0.005, Test_acc:97.0%, Test_loss:0.079
Epoch:12, Train_acc:99.6%, Train_loss:0.014, Test_acc:96.5%, Test_loss:0.085
Epoch:13, Train_acc:100.0%, Train_loss:0.005, Test_acc:97.0%, Test_loss:0.080
Epoch:14, Train_acc:100.0%, Train_loss:0.004, Test_acc:97.0%, Test_loss:0.065
Epoch:15, Train_acc:100.0%, Train_loss:0.007, Test_acc:95.8%, Test_loss:0.107
Epoch:16, Train_acc:99.9%, Train_loss:0.005, Test_acc:97.0%, Test_loss:0.065
Epoch:17, Train_acc:100.0%, Train_loss:0.003, Test_acc:96.5%, Test_loss:0.071
Epoch:18, Train_acc:99.9%, Train_loss:0.005, Test_acc:96.7%, Test_loss:0.081
Epoch:19, Train_acc:100.0%, Train_loss:0.003, Test_acc:96.5%, Test_loss:0.080
Epoch:20, Train_acc:99.9%, Train_loss:0.006, Test_acc:96.5%, Test_loss:0.078
Done
589
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
