```python # -*- coding: utf-8 -*- """ 🐱🐶 猫狗图像二分类（基于 PyTorch CPU） ✅ 使用文件夹结构数据集：train/ 和 test/ ✅ 基于卷积神经网络（CNN）完成分类 ✅ 输出测试准确率并绘制训练过程线图 ✅ 完整、可运行、无需额外配置 """ # Step 1: 安装依赖（首次运行时启用） # !pip install torch torchvision matplotlib pillow numpy # Step 2: 导入库 import os import torch import torch.nn as nn import torch.optim as optim from torch.utils.data import DataLoader from torchvision import transforms, datasets import matplotlib.pyplot as plt # 设置设备为 CPU（显式声明） device = torch.device('cpu') print(f"使用设备: {device}") # 设置随机种子以提高可复现性 torch.manual_seed(42) # ============================================= # 参数设置 # ============================================= data_dir = './' # 数据集根目录，包含 train/ 和 test/ 文件夹 batch_size = 32 num_epochs = 10 lr = 0.001 # 图像预处理 transform = transforms.Compose([ transforms.Resize((128, 128)), # 统一分辨率 transforms.Grayscale(num_output_channels=1), # 转为单通道灰度图 transforms.ToTensor(), # 转为张量 transforms.Normalize(mean=[0.5], std=[0.5]) # 归一化到 [-1, 1] ]) # ============================================= # 加载训练集和测试集 # ============================================= train_dataset = datasets.ImageFolder(root=os.path.join(data_dir, 'train'), transform=transform) test_dataset = datasets.ImageFolder(root=os.path.join(data_dir, 'test'), transform=transform) train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True) test_loader = DataLoader(test_dataset, batch_size=batch_size, shuffle=False) print(f"✅ 训练样本数: {len(train_dataset)}") print(f"✅ 测试样本数: {len(test_dataset)}") print(f"✅ 类别名称: {train_dataset.classes}") # 应为 ['cat', 'dog'] # ============================================= # 定义 CNN 模型 # ============================================= class CatDogCNN(nn.Module): def __init__(self): super(CatDogCNN, self).__init__() self.network = nn.Sequential( nn.Conv2d(1, 16, kernel_size=3, padding=1), # 输入单通道 nn.ReLU(), nn.MaxPool2d(2), nn.Conv2d(16, 32, kernel_size=3, padding=1), nn.ReLU(), nn.MaxPool2d(2), nn.Conv2d(32, 64, kernel_size=3, padding=1), nn.ReLU(), nn.MaxPool2d(2), nn.AdaptiveAvgPool2d((8, 8)), # 固定输出大小 nn.Flatten(), nn.Linear(64 * 8 * 8, 128), nn.ReLU(), nn.Dropout(0.5), nn.Linear(128, 2) # 两类输出 ) def forward(self, x): return self.network(x) model = CatDogCNN().to(device) # 损失函数和优化器 criterion = nn.CrossEntropyLoss() optimizer = optim.Adam(model.parameters(), lr=lr) # ============================================= # 训练模型 # ============================================= train_losses = [] train_accuracies = [] test_accuracies = [] print("\n🚀 开始训练...") for epoch in range(num_epochs): # 训练阶段 model.train() running_loss = 0.0 correct = 0 total = 0 for images, labels in train_loader: images, labels = images.to(device), labels.to(device) outputs = model(images) loss = criterion(outputs, labels) optimizer.zero_grad() loss.backward() optimizer.step() running_loss += loss.item() _, predicted = torch.max(outputs, 1) total += labels.size(0) correct += (predicted == labels).sum().item() avg_loss = running_loss / len(train_loader) train_acc = correct / total train_losses.append(avg_loss) train_accuracies.append(train_acc) # 测试阶段 model.eval() test_correct = 0 test_total = 0 with torch.no_grad(): for images, labels in test_loader: images, labels = images.to(device), labels.to(device) outputs = model(images) _, predicted = torch.max(outputs, 1) test_total += labels.size(0) test_correct += (predicted == labels).sum().item() test_acc = test_correct / test_total test_accuracies.append(test_acc) print(f"Epoch [{epoch+1}/{num_epochs}], " f"Loss: ${avg_loss:.4f}$, " f"Train Acc: ${train_acc:.4f}$, " f"Test Acc: ${test_acc:.4f}$") # ============================================= # 输出最终准确率 # ============================================= final_test_acc = test_accuracies[-1] print(f"\n✅ 最终测试准确率: ${final_test_acc:.4f}$") # ============================================= # 绘制训练曲线 # ============================================= plt.figure(figsize=(12, 5)) # 子图1：训练损失 plt.subplot(1, 2, 1) plt.plot(range(1, num_epochs+1), train_losses, 'b-o', label='Training Loss') plt.title('训练损失曲线') plt.xlabel('Epoch') plt.ylabel('Loss') plt.grid(True) plt.legend() # 子图2：准确率对比 plt.subplot(1, 2, 2) epochs_range = range(1, num_epochs+1) plt.plot(epochs_range, train_accuracies, 'g-o', label='Train Accuracy') plt.plot(epochs_range, test_accuracies, 'r-o', label='Test Accuracy') plt.title('训练与测试准确率') plt.xlabel('Epoch') plt.ylabel('Accuracy') plt.ylim(0, 1) plt.grid(True) plt.legend() plt.tight_layout() plt.show() # 显示最终结果 print(f"\n📊 模型训练完成！") print(f"📈 最高测试准确率: ${max(test_accuracies):.4f}$") print(f"🎯 最终测试准确率: ${final_test_acc:.4f}$") ``` --- ```python # -*- coding: utf-8 -*- """ 🐱🐶 猫狗分类（ResNet-like 小型网络 + 更强数据增强） ✅ 使用更复杂的网络结构 ✅ 添加更多数据增强手段 ✅ 输出测试准确率与动态训练线图 ✅ 支持文件夹输入：train/, test/ """ # Step 1: 安装依赖（首次运行时启用） # !pip install torch torchvision matplotlib pillow # Step 2: 导入库 import os import torch import torch.nn as nn import torch.optim as optim from torch.utils.data import DataLoader from torchvision import transforms, datasets import matplotlib.pyplot as plt # 固定随机种子 torch.manual_seed(42) # ============================================= # 参数设置 # ============================================= data_dir = './' batch_size = 32 num_epochs = 10 lr = 0.001 device = torch.device('cpu') # 更强的数据增强（适用于训练集） train_transform = transforms.Compose([ transforms.Resize((128, 128)), transforms.Grayscale(num_output_channels=1), transforms.RandomHorizontalFlip(p=0.5), # 随机翻转 transforms.RandomRotation(degrees=10), # 随机旋转 transforms.ColorJitter(brightness=0.2), # 亮度扰动 transforms.ToTensor(), transforms.Normalize(mean=[0.5], std=[0.5]) ]) # 测试集仅做基本变换 test_transform = transforms.Compose([ transforms.Resize((128, 128)), transforms.Grayscale(num_output_channels=1), transforms.ToTensor(), transforms.Normalize(mean=[0.5], std=[0.5]) ]) # 加载数据集 train_dataset = datasets.ImageFolder(root=os.path.join(data_dir, 'train'), transform=train_transform) test_dataset = datasets.ImageFolder(root=os.path.join(data_dir, 'test'), transform=test_transform) train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True) test_loader = DataLoader(test_dataset, batch_size=batch_size, shuffle=False) print(f"✅ 训练集大小: {len(train_dataset)}") print(f"✅ 测试集大小: {len(test_dataset)}") print(f"✅ 类别: {train_dataset.classes}") # ============================================= # 定义改进型 CNN（类似 Residual 结构简化版） # ============================================= class SmallResNet(nn.Module): def __init__(self): super(SmallResNet, self).__init__() self.conv1 = nn.Sequential( nn.Conv2d(1, 16, 3, padding=1), nn.BatchNorm2d(16), nn.ReLU(), nn.MaxPool2d(2) ) self.conv2 = nn.Sequential( nn.Conv2d(16, 32, 3, padding=1), nn.BatchNorm2d(32), nn.ReLU(), nn.MaxPool2d(2) ) self.conv3 = nn.Sequential( nn.Conv2d(32, 64, 3, padding=1), nn.BatchNorm2d(64), nn.ReLU(), nn.MaxPool2d(2) ) self.identity = nn.Identity() self.global_pool = nn.AdaptiveAvgPool2d((4, 4)) self.classifier = nn.Sequential( nn.Flatten(), nn.Linear(64 * 4 * 4, 128), nn.ReLU(), nn.Dropout(0.5), nn.Linear(128, 2) ) def forward(self, x): x = self.conv1(x) x = self.conv2(x) x = self.conv3(x) x = self.global_pool(x) x = self.classifier(x) return x model = SmallResNet().to(device) criterion = nn.CrossEntropyLoss() optimizer = optim.Adam(model.parameters(), lr=lr) # ============================================= # 训练记录 # ============================================= losses = [] train_accs = [] test_accs = [] print("\n🚀 开始训练（增强模型）...") for epoch in range(num_epochs): model.train() loss_epoch = 0.0 correct_train = 0 total_train = 0 for x, y in train_loader: x, y = x.to(device), y.to(device) out = model(x) loss = criterion(out, y) optimizer.zero_grad() loss.backward() optimizer.step() loss_epoch += loss.item() _, pred = out.max(1) total_train += y.size(0) correct_train += (pred == y).sum().item() losses.append(loss_epoch / len(train_loader)) train_accs.append(correct_train / total_train) # 测试准确率 model.eval() correct_test = 0 total_test = 0 with torch.no_grad(): for x, y in test_loader: x, y = x.to(device), y.to(device) out = model(x) _, pred = out.max(1) total_test += y.size(0) correct_test += (pred == y).sum().item() test_accs.append(correct_test / total_test) print(f"Epoch {epoch+1}/{num_epochs} | Loss: ${losses[-1]:.4f}$ | " f"Train Acc: ${train_accs[-1]:.4f}$ | Test Acc: ${test_accs[-1]:.4f}$") # ============================================= # 输出最终结果 # ============================================= final_acc = test_accs[-1] print(f"\n✅ 最终测试准确率: ${final_acc:.4f}$") # ============================================= # 绘制训练线图 # ============================================= plt.figure(figsize=(10, 6)) plt.plot(losses, 'b-o', label='Training Loss', markersize=4) plt.plot(train_accs, 'g-o', label='Train Accuracy', markersize=4) plt.plot(test_accs, 'r-o', label='Test Accuracy', markersize=4) plt.title('模型训练过程（损失 & 准确率）') plt.xlabel('Epoch') plt.ylabel('Value') plt.ylim(0, 1.2) plt.legend() plt.grid(True) plt.tight_layout() plt.show() print(f"\n🎉 训练完成！最终测试准确率: ${final_acc:.4f}$") ``` --- ### ✅ 使用说明 #### 数据结构要求： ``` 当前目录/ │ ├── train/ │ ├── cat/ │ │ ├── cat1.jpg, cat2.png... │ └── dog/ │ ├── dog1.jpg, dog2.png... │ └── test/ ├── cat/ │ ├── cat_test1.jpg... └── dog/ ├── dog_test1.jpg... ``` #### 如何运行？ 1. 将上述任一代码保存为 `.ipynb` 或 `.py` 文件 2. 放入与 `train/` 和 `test/` 同级目录 3. 运行代码即可自动训练并出图 --- 知识点（列出该代码中遇到的知识点） 1. **卷积神经网络（CNN）**：通过卷积层自动提取图像局部特征，适用于图像分类任务。 2. **交叉熵损失函数**：衡量预测概率分布与真实标签差异，驱动模型参数更新。 3. **数据增强技术**：如翻转、旋转、亮度调整，提升模型泛化能力，防止过拟合。