pytorch实现自己的深度神经网络（公共数据集）

一、训练文件——train.py

注意：在运行此代码之前，需要配置好pytorch-GPU版本的环境，具体再次不谈。

import torch
import torch.nn as nn
import torch.optim as optim
import torchvision
import torchvision.transforms as transforms# 检查GPU是否可用
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
print("Device:", device)# 数据预处理的转换
transform = transforms.Compose([transforms.Resize((256, 256)),transforms.ToTensor(),transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])# 加载CIFAR-10训练数据集
train_dataset = torchvision.datasets.CIFAR10(root='./data', train=True,download=True, transform=transform)# 创建数据加载器
train_loader = torch.utils.data.DataLoader(train_dataset, batch_size=8,shuffle=True, num_workers=0)# 定义神经网络模型
class CNN(nn.Module):def __init__(self):super(CNN, self).__init__()self.conv1 = nn.Conv2d(3, 32, 3, padding=1)self.conv2 = nn.Conv2d(32, 64, 3, padding=1)self.conv3 = nn.Conv2d(64, 128, 3, padding=1)self.pool = nn.MaxPool2d(2, 2)self.fc1 = nn.Linear(128 * 32 * 32, 512)self.fc2 = nn.Linear(512, 10)def forward(self, x):x = self.pool(torch.relu(self.conv1(x)))x = self.pool(torch.relu(self.conv2(x)))x = self.pool(torch.relu(self.conv3(x)))x = x.view(-1, 128 * 32 * 32)x = torch.relu(self.fc1(x))x = self.fc2(x)return x# 实例化模型，并将其移动到可用设备上
model = CNN().to(device)# 定义损失函数
criterion = nn.CrossEntropyLoss()# 定义优化器
optimizer = optim.SGD(model.parameters(), lr=0.001, momentum=0.9)if __name__ == '__main__':# 训练神经网络for epoch in range(5):running_loss = 0.0for i, data in enumerate(train_loader, 0):inputs, labels = data[0].to(device), data[1].to(device)# 梯度清零optimizer.zero_grad()# 正向传播outputs = model(inputs)loss = criterion(outputs, labels)# 反向传播 + 优化loss.backward()optimizer.step()# 打印统计信息running_loss += loss.item()if i % 200 == 199:print('[%d, %5d] loss: %.3f' %(epoch + 1, i + 1, running_loss / 200))running_loss = 0.0print('Finished Training')# 保存模型至文件torch.save(model.state_dict(), 'cifar10_cnn_model.pth')

二、测试文件——val.py

import torch
import torch.nn as nn
import torch.optim as optim
import torchvision
import torchvision.transforms as transforms
import matplotlib.pyplot as plt
import numpy as np
import cv2# 检查GPU是否可用
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
print("Device:", device)# 数据预处理的转换
transform = transforms.Compose([transforms.Resize((256, 256)),transforms.ToTensor(),transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])# 加载CIFAR-10测试数据集
test_dataset = torchvision.datasets.CIFAR10(root='./data', train=False,download=True, transform=transform)# 创建测试数据加载器
test_loader = torch.utils.data.DataLoader(test_dataset, batch_size=8,shuffle=False, num_workers=0)# 加载模型并将其移动到可用设备上
class CNN(nn.Module):def __init__(self):super(CNN, self).__init__()self.conv1 = nn.Conv2d(3, 32, 3, padding=1)self.conv2 = nn.Conv2d(32, 64, 3, padding=1)self.conv3 = nn.Conv2d(64, 128, 3, padding=1)self.pool = nn.MaxPool2d(2, 2)self.fc1 = nn.Linear(128 * 32 * 32, 512)self.fc2 = nn.Linear(512, 10)def forward(self, x):x = self.pool(torch.relu(self.conv1(x)))x = self.pool(torch.relu(self.conv2(x)))x = self.pool(torch.relu(self.conv3(x)))x = x.view(-1, 128 * 32 * 32)x = torch.relu(self.fc1(x))x = self.fc2(x)return x
# 显示函数
def imshow(img):img = img / 2 + 0.5npimg = img.numpy()# 坐标转换plt.imshow(np.transpose(npimg, (1, 2, 0)))plt.show()model = CNN().to(device)
model.load_state_dict(torch.load('cifar10_cnn_model.pth'))
model.eval()if __name__ == '__main__':# 在测试集上测试模型correct = 0total = 0with torch.no_grad():for data in test_loader:images, labels = data[0].to(device), data[1].to(device)outputs = model(images)# 预测值的最大值以及最大值的类别索引_, predicted = torch.max(outputs, 1)total += labels.size(0)correct += (predicted == labels).sum().item()print('Accuracy on the test images: %d %%' % (100 * correct / total))# 显示测试集中的一些图片及其预测结果# 生成一个迭代器，从数据加载器中取出数据dataiter = iter(test_loader)# 从迭代器中获取下一个批次的数据images, labels = dataiter.next()# 将获取到的批次数据移动到device上，在这里也就是GPU上images, labels = images.to(device), labels.to(device)dip_flag = Falseif dip_flag == True:# -------------------------------------------# 可以选择 使用opencv显示# -------------------------------------------np_images = images.cpu().numpy()# 循环遍历并显示所有测试集图片for i in range(len(np_images)):# 从归一化中还原图像数据np_image = np.transpose(np_images[i], (1, 2, 0))   # 从CHW转换为HWCnp_image = np_image * 0.5 + 0.5# 将图像数据从float类型转换为unit8类型np_image = (np_image * 255).astype(np.uint8)# 使用opencv显示图像cv2.imshow("Image {}".format(i+1), np_image)cv2.waitKey(0)# 等待用户按下任意键继续显示下一张图像cv2.destroyAllWindows()imshow(torchvision.utils.make_grid(images.cpu()))print('GroundTruth: ', ' '.join('%5s' % test_dataset.classes[labels[j]] for j in range(8)))outputs = model(images)_, predicted = torch.max(outputs, 1)print('Predicted: ', ' '.join('%5s' % test_dataset.classes[predicted[j]]for j in range(8)))