【Pytorch深度学习实践】刘二大人11 Advanced CNN-CSDN博客

本文链接：https://blog.csdn.net/weixin_45757219/article/details/131639053

GoogleNet

定义网络时要注意减少代码冗余（eg:函数/类），找出相同的代码块
Inception块：
在这里插入图片描述

四条路算完要保证图像的高度和宽度要一致（b,c,w,h），c可以不同（做padding）
Average Pooling：设置相应的padding和stride 保证图像大小不变
Concatenate:把张量拼接在一起
括号里面的是输出通道数的大小

1*1卷积

个数取决于输入张量的通道数
在这里插入图片描述
最主要的工作是，改变通道数量 C1->C2

两种方式最终运算结果是一样的，但是运算量相差太大

初始通道数作为参数，可以进行实例化
在这里插入图片描述

定义网络时，把这两行去掉，让计算机去求。报错信息里面也会有

inception块实现手写数字识别

import torch
import torch.nn as nn
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
 
# prepare dataset
 
batch_size = 64
transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.1307,), (0.3081,))]) # 归一化,均值和方差
 
train_dataset = datasets.MNIST(root='../dataset/mnist/', train=True, download=True, transform=transform)
train_loader = DataLoader(train_dataset, shuffle=True, batch_size=batch_size)
test_dataset = datasets.MNIST(root='../dataset/mnist/', train=False, download=True, transform=transform)
test_loader = DataLoader(test_dataset, shuffle=False, batch_size=batch_size)
 
# design model using class
class InceptionA(nn.Module):
    def __init__(self, in_channels):
        super(InceptionA, self).__init__()
        self.branch1x1 = nn.Conv2d(in_channels, 16, kernel_size=1)#第二个分支 1*1分支
 
        self.branch5x5_1 = nn.Conv2d(in_channels, 16, kernel_size=1)#1*1卷积
        self.branch5x5_2 = nn.Conv2d(16, 24, kernel_size=5, padding=2)#第三个分支  5*5卷积 上一层输出通道数等于这一层的输入通道数
 
        self.branch3x3_1 = nn.Conv2d(in_channels, 16, kernel_size=1)
        self.branch3x3_2 = nn.Conv2d(16, 24, kernel_size=3, padding=1)
        self.branch3x3_3 = nn.Conv2d(24, 24, kernel_size=3, padding=1)#第四个分支 
 
        self.branch_pool = nn.Conv2d(in_channels, 24, kernel_size=1)#1*1卷积核 第一个分支
 
    def forward(self, x):
        branch1x1 = self.branch1x1(x)#直接将x送入到branch里面
 
        branch5x5 = self.branch5x5_1(x)
        branch5x5 = self.branch5x5_2(branch5x5)
 
        branch3x3 = self.branch3x3_1(x)
        branch3x3 = self.branch3x3_2(branch3x3)
        branch3x3 = self.branch3x3_3(branch3x3)
 
        branch_pool = F.avg_pool2d(x, kernel_size=3, stride=1, padding=1)#先平均池化   stride和padding保证图片大小不变
        branch_pool = self.branch_pool(branch_pool)#将池化分支的值进行1*1卷积

#四个分支的输出图片大小一样 输出通道数不一样
        outputs = [branch1x1, branch5x5, branch3x3, branch_pool]#一共输出88个通道
        return torch.cat(outputs, dim=1) #concatenate操作 沿dim=1的维度拼接在一起 即c (b,c,w,h) (0,1,2,3)
 
class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
        self.conv2 = nn.Conv2d(88, 20, kernel_size=5) # 88 = 24x3 + 16 由inception来的
 
        self.incep1 = InceptionA(in_channels=10) # 与conv1 中的10对应
        self.incep2 = InceptionA(in_channels=20) # 与conv2 中的20对应
 
        self.mp = nn.MaxPool2d(2)
        self.fc = nn.Linear(1408, 10) #1408 就是到incep2后的输出
 
 
    def forward(self, x):
        in_size = x.size(0)
        x = F.relu(self.mp(self.conv1(x)))#conve1后通道变成10
        x = self.incep1(x)#输出通道变成88个
        x = F.relu(self.mp(self.conv2(x)))#88->20
        x = self.incep2(x)#20->88
        x = x.view(in_size, -1)#变成向量
        x = self.fc(x)
 
        return x
 
model = Net()
 
# construct loss and optimizer
criterion = torch.nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(), lr=0.01, momentum=0.5)
 
# training cycle forward, backward, update
 
 
def train(epoch):
    running_loss = 0.0
    for batch_idx, data in enumerate(train_loader, 0):
        inputs, target = data
        optimizer.zero_grad()
 
        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():
    correct = 0
    total = 0
    with torch.no_grad():
        for data in test_loader:
            images, labels = data
            outputs = model(images)
            _, predicted = torch.max(outputs.data, dim=1)
            total += labels.size(0)
            correct += (predicted == labels).sum().item()
    print('accuracy on test set: %d %% ' % (100*correct/total))
 
 
if __name__ == '__main__':
    for epoch in range(10):
        train(epoch)
        test()

重点：怎样改变卷积层的结构提高性能
训练轮数不是越多越好，可能训练到一半网络已经到达最优了，观察test决定到哪轮数 epoch合适
某一次测试集的准确率达到最高点，就把当前网络的参数进行存盘

ResNet

在这里插入图片描述
梯度消失：梯度趋于0时，权重更新时，权重得不到什么更新

在这里插入图片描述
加锁解决梯度消失的问题。在深度学习里面比较难实现，因为层数太多了。
解决办法：跳跃连接

说明：1、要解决的问题：梯度消失
2、跳连接，H(x) = F(x) + x,前面两层的输出和x输入张量维度，通道，宽度，高度必须一样，加完后再激活。不要做pooling，张量的维度会发生变化。

为什么会解决：
在这里插入图片描述
求导加了1，
趋于0时，在1附近而不是在0附近，链式法则时就不会出现梯度消失的问题，因此可以对开始的层进行充分的训练

在这里插入图片描述
把网络超参数和size要算出来，写代码时要进行测试，不要一次性写完，看每一层网络输出和预测的是不是一样的

import torch
import torch.nn as nn
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
 
# prepare dataset
 
batch_size = 64
transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.1307,), (0.3081,))]) # 归一化,均值和方差
 
train_dataset = datasets.MNIST(root='../dataset/mnist/', train=True, download=True, transform=transform)
train_loader = DataLoader(train_dataset, shuffle=True, batch_size=batch_size)
test_dataset = datasets.MNIST(root='../dataset/mnist/', train=False, download=True, transform=transform)
test_loader = DataLoader(test_dataset, shuffle=False, batch_size=batch_size)
 
# design model using class
class ResidualBlock(nn.Module):
    def __init__(self, channels):
        super(ResidualBlock, self).__init__()
        self.channels = channels
        self.conv1 = nn.Conv2d(channels, channels, kernel_size=3, padding=1)#保证输入输出通道、大小不变
        self.conv2 = nn.Conv2d(channels, channels, kernel_size=3, padding=1)#保证输入输出通道、大小不变
 
    def forward(self, x):
        y = F.relu(self.conv1(x))
        y = self.conv2(y)
        return F.relu(x + y)
 
class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = nn.Conv2d(1, 16, kernel_size=5)
        self.conv2 = nn.Conv2d(16, 32, kernel_size=5) # 88 = 24x3 + 16
 
        self.rblock1 = ResidualBlock(16)#输入维度为16
        self.rblock2 = ResidualBlock(32)
 
        self.mp = nn.MaxPool2d(2)
        self.fc = nn.Linear(512, 10) # 暂时不知道1408咋能自动出来的
 
 
    def forward(self, x):
        in_size = x.size(0)
 
        x = self.mp(F.relu(self.conv1(x)))
        x = self.rblock1(x)
        x = self.mp(F.relu(self.conv2(x)))
        x = self.rblock2(x)
 
        x = x.view(in_size, -1)
        x = self.fc(x)
        return x
 
model = Net()
 
# construct loss and optimizer
criterion = torch.nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(), lr=0.01, momentum=0.5)
 
# training cycle forward, backward, update
 
 
def train(epoch):
    running_loss = 0.0
    for batch_idx, data in enumerate(train_loader, 0):
        inputs, target = data
        optimizer.zero_grad()
 
        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():
    correct = 0
    total = 0
    with torch.no_grad():
        for data in test_loader:
            images, labels = data
            outputs = model(images)
            _, predicted = torch.max(outputs.data, dim=1)
            total += labels.size(0)
            correct += (predicted == labels).sum().item()
    print('accuracy on test set: %d %% ' % (100*correct/total))
 
 
if __name__ == '__main__':
    for epoch in range(10):
        train(epoch)
        test()