• pytorch-构建卷积神经网络

    构建卷积神经网络

    • 卷积网络中的输入和层与传统神经网络有些区别，需重新设计，训练模块基本一致 
      1. import torch
      2. import torch.nn as nn
      3. import torch.optim as optim
      4. import torch.nn.functional as F
      5. from torchvision import datasets,transforms 
      6. import matplotlib.pyplot as plt
      7. import numpy as np
      8. %matplotlib inline

      首先读取数据

    • 分别构建训练集和测试集（验证集）
    • DataLoader来迭代取数据 
      1. # 定义超参数 
      2. input_size = 28  #图像的总尺寸28*28
      3. num_classes = 10  #标签的种类数
      4. num_epochs = 3  #训练的总循环周期
      5. batch_size = 64  #一个撮（批次）的大小，64张图片
      6.  
      7. # 训练集
      8. train_dataset = datasets.MNIST(root='./data',  
      9.                             train=True,   
      10.                             transform=transforms.ToTensor(),  
      11.                             download=True) 
      12.  
      13. # 测试集
      14. test_dataset = datasets.MNIST(root='./data', 
      15.                            train=False, 
      16.                            transform=transforms.ToTensor())
      17.  
      18. # 构建batch数据
      19. train_loader = torch.utils.data.DataLoader(dataset=train_dataset, 
      20.                                            batch_size=batch_size, 
      21.                                            shuffle=True)
      22. test_loader = torch.utils.data.DataLoader(dataset=test_dataset, 
      23.                                            batch_size=batch_size, 
      24.                                            shuffle=True)

      卷积网络模块构建

    • 一般卷积层，relu层，池化层可以写成一个套餐
    • 注意卷积最后结果还是一个特征图，需要把图转换成向量才能做分类或者回归任务 
      1. class CNN(nn.Module):
      2.     def __init__(self):
      3.         super(CNN, self).__init__()
      4.         self.conv1 = nn.Sequential(         # 输入大小 (1, 28, 28)
      5.             nn.Conv2d(
      6.                 in_channels=1,              # 灰度图
      7.                 out_channels=16,            # 要得到几多少个特征图
      8.                 kernel_size=5,              # 卷积核大小
      9.                 stride=1,                   # 步长
      10.                 padding=2,                  # 如果希望卷积后大小跟原来一样，需要设置padding=(kernel_size-1)/2 if stride=1
      11.             ),                              # 输出的特征图为 (16, 28, 28)
      12.             nn.ReLU(),                      # relu层
      13.             nn.MaxPool2d(kernel_size=2),    # 进行池化操作（2x2 区域）, 输出结果为： (16, 14, 14)
      14.         )
      15.         self.conv2 = nn.Sequential(         # 下一个套餐的输入 (16, 14, 14)
      16.             nn.Conv2d(16, 32, 5, 1, 2),     # 输出 (32, 14, 14)
      17.             nn.ReLU(),                      # relu层
      18.             nn.Conv2d(32, 32, 5, 1, 2),
      19.             nn.ReLU(),
      20.             nn.MaxPool2d(2),                # 输出 (32, 7, 7)
      21.         )
      22.         
      23.         self.conv3 = nn.Sequential(         # 下一个套餐的输入 (16, 14, 14)
      24.             nn.Conv2d(32, 64, 5, 1, 2),     # 输出 (32, 14, 14)
      25.             nn.ReLU(),             # 输出 (32, 7, 7)
      26.         )
      27.         
      28.         self.out = nn.Linear(64 * 7 * 7, 10)   # 全连接层得到的结果
      29.  
      30.     def forward(self, x):
      31.         x = self.conv1(x)
      32.         x = self.conv2(x)
      33.         x = self.conv3(x)
      34.         x = x.view(x.size(0), -1)           # flatten操作，结果为：(batch_size, 32 * 7 * 7)
      35.         output = self.out(x)
      36.         return output

      准确率作为评估标准

      1. def accuracy(predictions, labels):
      2.     pred = torch.max(predictions.data, 1)[1] 
      3.     rights = pred.eq(labels.data.view_as(pred)).sum() 
      4.     return rights, len(labels)

      训练网络模型

      1. # 实例化
      2. net = CNN() 
      3. #损失函数
      4. criterion = nn.CrossEntropyLoss() 
      5. #优化器
      6. optimizer = optim.Adam(net.parameters(), lr=0.001) #定义优化器，普通的随机梯度下降算法
      7.  
      8. #开始训练循环
      9. for epoch in range(num_epochs):
      10.     #当前epoch的结果保存下来
      11.     train_rights = [] 
      12.     
      13.     for batch_idx, (data, target) in enumerate(train_loader):  #针对容器中的每一个批进行循环
      14.         net.train()                             
      15.         output = net(data) 
      16.         loss = criterion(output, target) 
      17.         optimizer.zero_grad() 
      18.         loss.backward() 
      19.         optimizer.step() 
      20.         right = accuracy(output, target) 
      21.         train_rights.append(right) 
      22.  
      23.     
      24.         if batch_idx % 100 == 0: 
      25.             
      26.             net.eval() 
      27.             val_rights = [] 
      28.             
      29.             for (data, target) in test_loader:
      30.                 output = net(data) 
      31.                 right = accuracy(output, target) 
      32.                 val_rights.append(right)
      33.                 
      34.             #准确率计算
      35.             train_r = (sum([tup[0] for tup in train_rights]), sum([tup[1] for tup in train_rights]))
      36.             val_r = (sum([tup[0] for tup in val_rights]), sum([tup[1] for tup in val_rights]))
      37.  
      38.             print('当前epoch: {} [{}/{} ({:.0f}%)]\t损失: {:.6f}\t训练集准确率: {:.2f}%\t测试集正确率: {:.2f}%'.format(
      39.                 epoch, batch_idx * batch_size, len(train_loader.dataset),
      40.                 100. * batch_idx / len(train_loader), 
      41.                 loss.data, 
      42.                 100. * train_r[0].numpy() / train_r[1], 
      43.                 100. * val_r[0].numpy() / val_r[1]))
      当前epoch: 0 [0/60000 (0%)]	损失: 2.300918	训练集准确率: 10.94%	测试集正确率: 10.10%
当前epoch: 0 [6400/60000 (11%)]	损失: 0.204191	训练集准确率: 78.06%	测试集正确率: 93.31%
当前epoch: 0 [12800/60000 (21%)]	损失: 0.039503	训练集准确率: 86.51%	测试集正确率: 96.69%
当前epoch: 0 [19200/60000 (32%)]	损失: 0.057866	训练集准确率: 89.93%	测试集正确率: 97.54%
当前epoch: 0 [25600/60000 (43%)]	损失: 0.069566	训练集准确率: 91.68%	测试集正确率: 97.68%
当前epoch: 0 [32000/60000 (53%)]	损失: 0.228793	训练集准确率: 92.85%	测试集正确率: 98.18%
当前epoch: 0 [38400/60000 (64%)]	损失: 0.111003	训练集准确率: 93.72%	测试集正确率: 98.16%
当前epoch: 0 [44800/60000 (75%)]	损失: 0.110226	训练集准确率: 94.28%	测试集正确率: 98.44%
当前epoch: 0 [51200/60000 (85%)]	损失: 0.014538	训练集准确率: 94.78%	测试集正确率: 98.60%
当前epoch: 0 [57600/60000 (96%)]	损失: 0.051019	训练集准确率: 95.14%	测试集正确率: 98.45%
当前epoch: 1 [0/60000 (0%)]	损失: 0.036383	训练集准确率: 98.44%	测试集正确率: 98.68%
当前epoch: 1 [6400/60000 (11%)]	损失: 0.088116	训练集准确率: 98.50%	测试集正确率: 98.37%
当前epoch: 1 [12800/60000 (21%)]	损失: 0.120306	训练集准确率: 98.59%	测试集正确率: 98.97%
当前epoch: 1 [19200/60000 (32%)]	损失: 0.030676	训练集准确率: 98.63%	测试集正确率: 98.83%
当前epoch: 1 [25600/60000 (43%)]	损失: 0.068475	训练集准确率: 98.59%	测试集正确率: 98.87%
当前epoch: 1 [32000/60000 (53%)]	损失: 0.033244	训练集准确率: 98.62%	测试集正确率: 99.03%
当前epoch: 1 [38400/60000 (64%)]	损失: 0.024162	训练集准确率: 98.67%	测试集正确率: 98.81%
当前epoch: 1 [44800/60000 (75%)]	损失: 0.006713	训练集准确率: 98.69%	测试集正确率: 98.17%
当前epoch: 1 [51200/60000 (85%)]	损失: 0.009284	训练集准确率: 98.69%	测试集正确率: 98.97%
当前epoch: 1 [57600/60000 (96%)]	损失: 0.036536	训练集准确率: 98.68%	测试集正确率: 98.97%
当前epoch: 2 [0/60000 (0%)]	损失: 0.125235	训练集准确率: 98.44%	测试集正确率: 98.73%
当前epoch: 2 [6400/60000 (11%)]	损失: 0.028075	训练集准确率: 99.13%	测试集正确率: 99.17%
当前epoch: 2 [12800/60000 (21%)]	损失: 0.029663	训练集准确率: 99.26%	测试集正确率: 98.39%
当前epoch: 2 [19200/60000 (32%)]	损失: 0.073855	训练集准确率: 99.20%	测试集正确率: 98.81%
当前epoch: 2 [25600/60000 (43%)]	损失: 0.018130	训练集准确率: 99.16%	测试集正确率: 99.09%
当前epoch: 2 [32000/60000 (53%)]	损失: 0.006968	训练集准确率: 99.15%	测试集正确率: 99.11%

  • 相关阅读:
    纯java代码实现输入框和框并对输入txt进行处理
    4WRBA6EB15-2X/G24N9Z4/M比例换向阀控制器
    SqlSessionFactory与SqlSession
    element-ui el-table-column 宽度不能动态设置问题
    linux进程控制类命令
    如何用wireshark过滤媒体流
    嵌入式IDE(2)：KEIL中SCF分散加载链接文件详解和实例分析
    rmq nameserver
    Python-数据结构-字典
    Vue3入门
  • 原文地址：https://blog.csdn.net/qq_65838372/article/details/132725190