PyTorch深度学习实践——卷积神经网络(基础篇)

参考资料

参考资料1:https://blog.csdn.net/bit452/article/details/109690712
参考资料2：http://biranda.top/Pytorch%E5%AD%A6%E4%B9%A0%E7%AC%94%E8%AE%B0011%E2%80%94%E2%80%94Simple_CNN/

卷积神经网络(基础篇)

1、conv

import torch
in_channels, out_channels = 5, 10#输入通道是5，输出通道是10
width, height = 100, 100

kernel_size = 3 #默认转为3*3，最好用奇数正方形

#在pytorch中的数据处理都是通过batch来实现的
#因此对于C*W*H的三个维度图像，在代码中实际上是一个B（batch）*C*W*H的四个维度的图像
batch_size = 1 # 1

#生成一个四维的随机数
input = torch.randn(batch_size, in_channels, width, height)

#Conv2d需要设定，输入输出的通道数以及卷积核尺寸
conv_layer = torch.nn.Conv2d(in_channels, out_channels, kernel_size=kernel_size)

output = conv_layer(input)

print(input.shape)# 输入的图像的形状
print(output.shape)#输出的图像的形状
print(conv_layer.weight.shape)#卷积核的形状
#torch.Size([1, 5, 100, 100])
#torch.Size([1, 10, 98, 98])
#torch.Size([10, 5, 3, 3]) #10就是输出的通道，5是输入的通道，3和3是卷积核的大小

2、max_pooling

import torch

input = [3,4,6,5,7,
         2,4,6,8,2,
         1,6,7,8,4,
         9,7,4,6,2,
         3,7,5,4,1]

#将输入变为B*C*W*H
input = torch.Tensor(input).view(1, 1, 5, 5)
#构建卷积层，输入是1个通道，输出是1个通道，卷积核3×3，padding设为1，偏置量bias设为false,stride=2可设置卷积核的步长为2，不写默认为1
conv_layer = torch.nn.Conv2d(1, 1, kernel_size=3, stride=2, padding=1, bias=False)
#构建卷积核，将卷积核改变形状为 输出通道数*输入通道数*W*H
kernel = torch.Tensor([1,2,3,4,5,6,7,8,9]).view(1, 1, 3, 3)
#将做出来的卷积核张量，赋值给卷积运算中的权重（参与卷积计算）
conv_layer.weight.data = kernel.data
output = conv_layer(input)
print(output)

3、kernal

import torch
input = [3,4,6,5,
         2,4,6,8,
         1,6,7,8,
         9,7,4,6]
#batch_size, in_channels, width, height
input = torch.Tensor(input).view(1, 1, 4, 4)

#kernel_size=2 则MaxPooling中的Stride也为2
maxpooling_layer = torch.nn.MaxPool2d(kernel_size=2)

output = maxpooling_layer(input)

print(output)

4、简单卷积神经网络的实现

import torch
from torchvision import transforms
from torchvision import datasets
from torch.utils.data import DataLoader
import torch.nn.functional as F
import torch.optim as optim

batch_size = 64
transform = transforms.Compose([
    transforms.ToTensor(),
    transforms.Normalize((0.1307,), (0.3081,))
    # 均值 和标准差，，为了能把图像数据映射成  (0,1)分布
])
#训练集
train_dataset = datasets.MNIST(root='../dataset/mnist',
                               train=True,
                               download=True,
                               transform=transform)
train_loader = DataLoader(train_dataset, shuffle=True, batch_size=batch_size)
#测试集
test_dataset = datasets.MNIST(root='../dataset/mnist',
                              train=False,
                              download=True,
                              transform=transform)
test_loader = DataLoader(dataset=test_dataset,batch_size=batch_size,shuffle=False,)

# 构建网络
class Net(torch.nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        #卷积层 输入通道1个，输出通道10个，核是5×5
        self.conv1 = torch.nn.Conv2d(1, 10, kernel_size=5)
        self.conv2 = torch.nn.Conv2d(10, 20, kernel_size=5)
        #池化层
        self.pooling = torch.nn.MaxPool2d(2)
        #全连接层
        self.fc = torch.nn.Linear(320, 10)

    def forward(self, x):
        # data from (n,1,28,28) to (n,784) ,n是batch
        batch_size = x.size(0)  # 这一步在干什么 ，得到样本数量n
        x = F.relu(self.pooling(self.conv1(x)))
        x = F.relu(self.pooling(self.conv2(x)))
        x = x.view(batch_size, 320)
        x = self.fc(x) # 交叉熵损失函数，最后一层没有激活函数，请注意
        return x
model = Net()
#cuda 0是选择第一块显卡，cuda 1是选择第二块显卡。
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
#将来model迁移到device
model.to(device)
#交叉熵损失
criterion = torch.nn.CrossEntropyLoss()
#梯度下降法，每一次迭代计算mini-batch
optimizer = torch.optim.SGD(model.parameters(), lr=0.01)

# 训练 —— 把一轮循环 封装成函数
def train(epoch):
    running_loss = 0.0
    for batch_idx, data in enumerate(train_loader, 0):
        inputs, target = data
        #send the inputs and targets at every step to the GPU。
        inputs, target = inputs.to(device), target.to(device)
        optimizer.zero_grad()# 优化器梯度清零

        # 前馈，反馈，更新
        outputs = model(inputs)
        loss = criterion(outputs, target)
        loss.backward()
        optimizer.step()

        running_loss += loss.item()
        if batch_idx % 300 == 299:
            print('[%d, %5d] loss: %.3f' %(epoch + 1, batch_idx + 1, running_loss/300))
            running_loss = 0.0

def test():
    corret = 0
    total = 0
    with torch.no_grad():   # 测试集不需要计算梯度
        for data in test_loader:
            inputs, target = data
            # send the inputs and targets at every step to the GPU。
            inputs, target = inputs.to(device), target.to(device)
            outputs = model(inputs)
            _, predicted = torch.max(outputs.data, dim=1) # 返回两个值，一个是最大值，一个最大值下标
            total += target.size(0)
            corret += (predicted == target).sum().item()
        print('Accuracy on test set: %d %%' %(100*corret / total))

if __name__=='__main__':
    for epoch in range(10):
        train(epoch)
        test()