pytorch学习笔记-高阶篇（卷积神经网络实战1）

本篇主要是卷积神经网络的实战，数据集用的是CIFAR-10

一、CIFAR-10介绍

  CIFAR-10有10大类图片，分别是：airplane，automobile，bird，cat，deer，dog，frog，horse，ship，truck。CIFAR-100是把这10大类细分为100类。每张图片的尺寸是32x32，每一类有6000张图片，一共60000张，按照5:1分为训练和测试数据集。

二、Lenet类

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


class Lenet5(nn.Module):
    """
    for cifar10 dataset
    """

    def __init__(self):
        super(Lenet5, self).__init__()

        self.conv_unit = nn.Sequential(
            # x:[b, 3, 32, 32] => [b, 6, ]
            nn.Conv2d(in_channels=3, out_channels=6, kernel_size=5, stride=1, padding=0),
            nn.AvgPool2d(kernel_size=2, stride=2, padding=0),
            #
            nn.Conv2d(in_channels=6, out_channels=16, kernel_size=5, stride=1, padding=0),
            nn.AvgPool2d(kernel_size=2, stride=2, padding=0),
            #
        )
        # flatten
        # full_connected unit
        self.fc_unit = nn.Sequential(
            nn.Linear(16 * 5 * 5, 120),
            nn.ReLU(),
            nn.Linear(120, 84),
            nn.ReLU(),
            nn.Linear(84, 10)
        )

        '''
        # [b, 3, 32, 32]
        # 全连接层的输入维度不知道，这里测试一下
        tmp = torch.randn(2, 3, 32, 32)
        
        out = self.conv_unit(tmp)
        print('conv out:', out.shape)
        # conv out: torch.Size([2, 16, 5, 5])
        '''

        # use Cross Entropy Loss
        # 分类问题用交叉熵更合适
        # self.criteon = nn.CrossEntropyLoss()

    def forward(self, x):
        """
        Args:
            x: [b, 3, 32, 32]
        Returns:
        """
        batchsz = x.size(0)
        # [b, 3, 32, 23] => [b, 16, 5, 5]
        x = self.conv_unit(x)
        # [b, 16, 5, 5] => [b, 16*5*5]
        x = x.view(batchsz, 16 * 5 * 5)
        # [b, 16*5*5] => [b, 10]
        # softmax/sigmoid前面的变量一般称为logits
        logits = self.fc_unit(x)

        # [b, 10]
        # 做Cross Entropy Loss里面包括了sofmax操作，不需要重复写
        # pred = F.softmax(logits, dim=1)
        # loss = self.criteon(logits, y)
        return logits


def main():
    net = Lenet5()
    tmp = torch.randn(2, 3, 32, 32)

    out = net(tmp)
    print('lenet out:', out.shape)


if __name__ == '__main__':
    main()

三、main

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

from lenet5 import Lenet5


def main():
    batchsz = 8
    # 训练数据集加载
    # 一次加载一张
    cifar_train = datasets.CIFAR10("./dataset", train=True, transform=transforms.Compose([
        transforms.Resize([32, 32]),
        transforms.ToTensor()
    ]), download=True)
    # 一次加载多张
    cifar_train = DataLoader(dataset=cifar_train, batch_size=batchsz, shuffle=True)

    # 测试数据集加载
    # 一次加载一张
    cifar_test = datasets.CIFAR10("./dataset", train=False, transform=transforms.Compose([
        transforms.Resize([32, 32]),
        transforms.ToTensor()
    ]), download=True)
    # 一次加载多张
    cifar_test = DataLoader(dataset=cifar_test, batch_size=batchsz, shuffle=True)

    # 迭代器
    x, label = iter(cifar_train).next()
    print('x:', x.shape, 'label:', label.shape)

    device = torch.device('cuda')
    model =Lenet5().to(device)
    criteon = nn.CrossEntropyLoss()
    optimizer = optim.Adam(model.parameters(), lr=1e-3)
    print(model)

    for epoch in range(1000):
        model.train()
        for batchidx, (x, label) in enumerate(cifar_train):
            # [b, 3, 32, 32]
            # [b]
            x, label = x.to(device), label.to(device)
            logits = model(x)
            # logits: [b, 10]
            # label: [b]
            # loss: tensor scalar
            loss = criteon(logits, label)

            # backprop
            optimizer.zero_grad()
            loss.backward()
            optimizer.step()

        #
        print(epoch, loss.item())

        model.eval()
        with torch.no_grad():
            # test
            total_correct = 0
            total_num = 0
            for x, label in cifar_test:
                # [b, 3, 32, 32]
                # [b]
                x, label = x.to(device), label.to(device)
                # [b, 10]
                logits = model(x)
                # [b]
                pred = logits.argmax(dim=1)
                # [b] vs [b] => scalar tensor
                total_correct = total_correct + torch.eq(pred, label).float().sum().item()
                total_num = total_num + x.size(0)

            acc = total_correct / total_num
            print(epoch, acc)


if __name__ == '__main__':
    main()