【深度学习05】 CIFAR10图像分类

⭐CIFAR-10是一个用于识别普适物体的小型数据集，它包含了10个类别的RGB彩色图片。图片尺寸：32 x 32。训练图片50000张，测试图片10000张

训练30轮后，正确率为66%，估计是网络结构太简单了，使用resnet-18训练10轮后，正确率可达80%

import torch
import torch.nn as nn
from torch.nn import Conv2d,MaxPool2d,Sequential,Flatten,Linear
import torchvision
from torch.utils.data import DataLoader
import time
from torch.utils.tensorboard import SummaryWriter

# 定义训练的设备
device = torch.device("cuda"if torch.cuda.is_available() else "cpu")
print(device)

# 下载训练集和测试集
train_data = torchvision.datasets.CIFAR10(root="MyLearn/data",train=True,transform=torchvision.transforms.ToTensor(),download=True)
test_data = torchvision.datasets.CIFAR10(root="MyLearn/data",train=False,transform=torchvision.transforms.ToTensor(),download=True)

# 计算长度
train_data_size = len(train_data)
test_data_size = len(test_data)
print("训练集数据长度为：{}".format(train_data_size))
print("测试集数据长度为：{}".format(test_data_size))

# 加载数据集
train_dataloader = DataLoader(train_data,batch_size=64)
test_dataloader = DataLoader(test_data,batch_size=64)

# 搭建神经网络
class zf_model(nn.Module):
    def __init__(self) -> None:
        super(zf_model,self).__init__()
        # 定义模型
        self.model1 = Sequential(
            Conv2d(3,32,5,1,2),
            MaxPool2d(2),
            Conv2d(32,32,5,1,2),
            MaxPool2d(2),
            Conv2d(32,64,5,1,2),
            MaxPool2d(2),
            Flatten(), # 平铺展开
            Linear(1024,64),
            Linear(64,10),
        )

    def forward(self,x):
        x = self.model1(x)

        return x

# 创建网络模型
zf = zf_model()
zf = zf.to(device) # 用GPU训练

# 损失函数：计算神经网络输出和正确答案的差距，loss越小越好
loss_fn = nn.CrossEntropyLoss() # 交叉熵损失
loss_fn = loss_fn.to(device)

# 优化器
learning_rate = 0.01 # 定义学习率
optimizer = torch.optim.SGD(zf.parameters(),lr=learning_rate) # 优化器：梯度下降法
# optimizer = torch.optim.Adam(zf.parameters(), lr=learning_rate)

# 设置训练网络的一些参数
# 记录训练的次数
total_train_step = 0
# 记录测试的次数
total_test_step = 0
# 训练的轮数
epoch = 35

# 添加tensorboard
writer = SummaryWriter("MyLearn/log_train")

# 计时开始
start_time = time.time()

for i in range(epoch):
    print("------第{}轮训练开始------".format(i+1))

    # 训练步骤开始
    for data in train_dataloader:
        imgs,targets = data # 读取图片和标签
        imgs = imgs.to(device)
        targets = targets.to(device)
        outputs = zf(imgs) # 让图片经过神经网络训练
        loss = loss_fn(outputs,targets) # 定义损失函数

        # 优化器优化模型
        # 使梯度趋近于零，即达到最低点
        optimizer.zero_grad()  # 清空梯度
        loss.backward()  # 反向传播，计算当前梯度
        optimizer.step()  # 一旦梯度被如backward()之类的函数计算好后，我们就可以调用这个函数

        total_train_step = total_train_step+1
        if total_train_step%100 == 0:
            end_time = time.time()
            print(end_time-start_time)
            # .item()：不考虑数据类型，只取出单纯的数字
            print("训练次数：{}，Loss：{}".format(total_train_step,loss.item()))
            writer.add_scalar("train_loss",loss.item(),total_train_step)

    # 测试步骤开始
    total_test_loss = 0
    total_accuracy = 0
    with torch.no_grad():
        for data in test_dataloader:
            imgs,targets = data
            imgs = imgs.to(device)
            targets = targets.to(device)
            outputs = zf(imgs)
            loss = loss_fn(outputs,targets)
            total_test_loss = total_test_loss+loss.item()
            accuracy = (outputs.argmax(1) == targets).sum()  # 计算正确率
            total_accuracy = total_accuracy+accuracy
    
    print("整体测试集上的Loss：{}".format(total_test_loss))
    print("整体测试集上的正确率：{}".format(total_accuracy/test_data_size))

    # 训练过程可视化
    writer.add_scalar("test_loss",total_test_loss,total_test_step)
    writer.add_scalar("test_accuracy",total_accuracy/test_data_size,total_test_step)
    total_test_step = total_test_step+1

    # 保存每一轮训练的模型
    # torch.save(zf,"path/zf.p()".format(i))
    # print("模型已保存")

writer.close()