LeNet

时间

1998

概念

代码

model

import torch.nn as nn
import torch.nn.functional as F


class LeNet(nn.Module):
    def __init__(self):
        super(LeNet, self).__init__()  # super()继承父类的构造函数
        self.conv1 = nn.Conv2d(3, 16, 5)
        self.pool1 = nn.MaxPool2d(2, 2)
        self.conv2 = nn.Conv2d(16, 32, 5)
        self.pool2 = nn.MaxPool2d(2, 2)
        self.fc1 = nn.Linear(32*5*5, 120)
        self.fc2 = nn.Linear(120, 84)
        self.fc3 = nn.Linear(84, 10)

    def forward(self, x): 
        x = F.relu(self.conv1(x))    # input(3, 32, 32) output(16, 28, 28)
        x = self.pool1(x)            # output(16, 14, 14)
        x = F.relu(self.conv2(x))    # output(32, 10, 10)
        x = self.pool2(x)            # output(32, 5, 5)
        x = x.view(-1, 32*5*5)       # output(32*5*5)
        x = F.relu(self.fc1(x))      # output(120)
        x = F.relu(self.fc2(x))      # output(84)
        x = self.fc3(x)              # output(10)
        return x

forward：定义正向传播的过程。

ReLU：激活函数

观察网络中的参数传递：发现传递的都是channel通道数，最后output在softmax函数里展开的也是展开的通道数。

train

import torch
import torchvision
import torch.nn as nn
from model import LeNet
import torch.optim as optim
import torchvision.transforms as transforms


def main():
    transform = transforms.Compose(
        [transforms.ToTensor(),
         transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])

    # 50000张训练图片
    # 第一次使用时要将download设置为True才会自动去下载数据集
    train_set = torchvision.datasets.CIFAR10(root='./data', train=True,
                                             download=True, transform=transform)
    train_loader = torch.utils.data.DataLoader(train_set, batch_size=36,
                                               shuffle=True, num_workers=0)

    # 10000张验证图片
    # 第一次使用时要将download设置为True才会自动去下载数据集
    val_set = torchvision.datasets.CIFAR10(root='./data', train=False,
                                           download=False, transform=transform)
    val_loader = torch.utils.data.DataLoader(val_set, batch_size=5000,
                                             shuffle=False, num_workers=0)
    val_data_iter = iter(val_loader)
    val_image, val_label = next(val_data_iter)
    
    # classes = ('plane', 'car', 'bird', 'cat',
    #            'deer', 'dog', 'frog', 'horse', 'ship', 'truck')

    net = LeNet()
    loss_function = nn.CrossEntropyLoss()
    optimizer = optim.Adam(net.parameters(), lr=0.001)

    for epoch in range(5):  # loop over the dataset multiple times

        running_loss = 0.0
        for step, data in enumerate(train_loader, start=0):
            # get the inputs; data is a list of [inputs, labels]
            inputs, labels = data

            # zero the parameter gradients
            optimizer.zero_grad()
            # forward + backward + optimize
            outputs = net(inputs)
            loss = loss_function(outputs, labels)
            loss.backward()
            optimizer.step()

            # print statistics
            running_loss += loss.item()
            if step % 500 == 499:    # print every 500 mini-batches
                with torch.no_grad():
                    outputs = net(val_image)  # [batch, 10]
                    predict_y = torch.max(outputs, dim=1)[1] # 
                    accuracy = torch.eq(predict_y, val_label).sum().item() / val_label.size(0)

                    print('[%d, %5d] train_loss: %.3f  test_accuracy: %.3f' %
                          (epoch + 1, step + 1, running_loss / 500, accuracy))
                    running_loss = 0.0

    print('Finished Training')

    save_path = './Lenet.pth'
    torch.save(net.state_dict(), save_path)


if __name__ == '__main__':
    main()

 predict_y = torch.max(outputs, dim=1)[1]  ：

max返回的两个值 0是最大值，1是位置

dim = 1 ：在维度1上去找最大值，第零个维度对应的是batch

为什么没计算一个batch，就需要调用一次optimizer.zero_grad()

如果不清楚历史梯度，就会对计算的历史梯度进行累加（通过这个特性你能够变相实现一个很大batch数值的训练）

predict.py

import torch
import torchvision.transforms as transforms
from PIL import Image

from model import LeNet


def main():
    transform = transforms.Compose(
        [transforms.Resize((32, 32)),
         transforms.ToTensor(),
         transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])

    classes = ('plane', 'car', 'bird', 'cat',
               'deer', 'dog', 'frog', 'horse', 'ship', 'truck')

    net = LeNet()
    net.load_state_dict(torch.load('Lenet.pth'))

    im = Image.open('1.jpg').convert('RGB')
    im = transform(im)  # [C, H, W]
    im = torch.unsqueeze(im, dim=0)  # [N, C, H, W]

    with torch.no_grad():
        outputs = net(im)
        predict = torch.max(outputs, dim=1)[1].numpy()
        # predict = torch.softmax(outputs,dim=1)
        # print(predict)
        # tensor([[9.9884e-01, 1.9386e-04, 3.8757e-04, 2.0671e-05, 2.5372e-04, 3.6199e-05,
        # 3.7643e-05, 1.7624e-04, 2.0138e-05, 3.4801e-05]])
    print(classes[int(predict)])


if __name__ == '__main__':
    main()

知识点：

增加新的维度： 

im = torch.unsqueeze(im, dim=0)  # [N, C, H, W] 

predict = torch.max(outputs, dim=1)[1].numpy()：

这一行代码使用torch.max()函数找到outputs张量在第一个维度上的最大值，并返回最大值和对应的索引。dim=1表示在第一个维度上进行最大值的计算，即对每个样本的输出进行比较。[1]表示返回最大值对应的索引。最后，.numpy()将结果转换为NumPy数组。 

更换：

predict = torch.softmax(outputs,dim=1)

print：tensor([[9.9884e-01, 1.9386e-04, 3.8757e-04, 2.0671e-05, 2.5372e-04, 3.6199e-05,
         3.7643e-05, 1.7624e-04, 2.0138e-05, 3.4801e-05]])

Pytorch使用