009.ResNet-FashionMNIST-正确率93.739

一、ResNet简介

ResNet是一次CNN网络架构，核心思想是引入"残差学习"来解决深层网络难以训练的问题。
在传统的网络中，每一层都直接尝试学习目标映射。相反，ResNet通过跨层连接，允许某一层学习输入与输出之间的残差（或者说是差异），使得这些网络层只需要学习与输入的微小差异，从而简化了学习目标和过程。

二、FashionMNIST数据集简介

之前的博客已经较为细致的介绍了FashionMNIST数据集：插眼传送

注意：了解数据集是机器学习的所有环节中最重要的一步，没有之一。

三、用代码实现FashionMNIST预测

1.导包

from torchvision.datasets import FashionMNIST
from torchvision.transforms import transforms
from torch.utils.data import DataLoader
import matplotlib.pyplot as plt
import torch
import numpy as np
import random

2.加载数据

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
#device = 'cpu'
generator = torch.Generator()# 设置随机种子，确保实验可重复性
seed_value = 420
torch.manual_seed(seed_value)
random.seed(seed_value)
np.random.seed(seed_value)
# 如果你使用CUDA并希望进一步确定性，可以添加下面两行代码
torch.cuda.manual_seed(seed_value)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
generator.manual_seed(seed_value)transform = transforms.Compose([transforms.RandomHorizontalFlip(), transforms.RandomRotation([-8,8]),transforms.ToTensor(),transforms.Normalize((0.5,), (0.5,))  # 归一化处理
])transform2 = transforms.Compose([#transforms.RandomRotation([-5,5]),transforms.ToTensor(),transforms.Normalize((0.5,), (0.5,))  # 归一化处理
])# 从"./dataset/"目录加载FashionMNIST数据集，如果没有则会自动下载。
train_data = FashionMNIST(root='./dataset/', train=True,  download=True,transform=transform)
test_data = FashionMNIST(root='./dataset/', train=False,  download=True,transform=transform2)
train_batch = DataLoader(dataset=train_data, batch_size=128,  shuffle=True, num_workers=0, drop_last=False, generator=generator)
test_batch = DataLoader(dataset=test_data, batch_size=128,  shuffle=False, num_workers=0, drop_last=False, generator=generator)

3.定义模型

class Model(torch.nn.Module):def __init__(self,in_features=1,out_features=10):super().__init__()self.relu = torch.nn.ReLU()self.conv1 = torch.nn.Conv2d(in_channels=in_features, out_channels=64, kernel_size=3, bias=False) self.adavgpool = torch.nn.AdaptiveAvgPool2d((1, 1))self.block1 = torch.nn.Sequential(self.conv1, torch.nn.BatchNorm2d(64), self.relu)self.output = torch.nn.Linear(512, out_features, bias=True)self.maxpool = torch.nn.AvgPool2d(2,ceil_mode=True)self.downsample = torch.nn.Sequential(torch.nn.Conv2d(64, 128, kernel_size=1,stride=2,bias=False),torch.nn.BatchNorm2d(128))self.downsample2 = torch.nn.Sequential(torch.nn.Conv2d(128, 256, kernel_size=1,stride=2, bias=False), torch.nn.BatchNorm2d(256))self.downsample3 = torch.nn.Sequential(torch.nn.Conv2d(256, 512, kernel_size=1,stride=2, bias=False), torch.nn.BatchNorm2d(512))self.conv_res = torch.nn.Sequential(torch.nn.Conv2d(in_channels=64, out_channels=128, kernel_size=3, padding=1, stride=2, bias=False),torch.nn.BatchNorm2d(128),self.relu,torch.nn.Conv2d(in_channels=128, out_channels=128, kernel_size=3, padding=1, bias=False),torch.nn.BatchNorm2d(128),)self.conv_res2 = torch.nn.Sequential(torch.nn.Conv2d(in_channels=128, out_channels=128, kernel_size=3, padding=1, stride=1, bias=False),torch.nn.BatchNorm2d(128),self.relu,torch.nn.Conv2d(in_channels=128, out_channels=128, kernel_size=3, padding=1, bias=False),torch.nn.BatchNorm2d(128),)self.conv_res3 = torch.nn.Sequential(torch.nn.Conv2d(in_channels=128, out_channels=256, kernel_size=3, padding=1, stride=2, bias=False),torch.nn.BatchNorm2d(256),self.relu,torch.nn.Conv2d(in_channels=256, out_channels=256, kernel_size=3, padding=1, bias=False),torch.nn.BatchNorm2d(256),)self.conv_res4 = torch.nn.Sequential(torch.nn.Conv2d(in_channels=256, out_channels=256, kernel_size=3, padding=1, stride=1, bias=False),torch.nn.BatchNorm2d(256),self.relu,torch.nn.Conv2d(in_channels=256, out_channels=256, kernel_size=3, padding=1, bias=False),torch.nn.BatchNorm2d(256),)self.conv_res5 = torch.nn.Sequential(torch.nn.Conv2d(in_channels=256, out_channels=512, kernel_size=3, padding=1, stride=2, bias=False),torch.nn.BatchNorm2d(512),self.relu,torch.nn.Conv2d(in_channels=512, out_channels=512, kernel_size=3, padding=1, bias=False),torch.nn.BatchNorm2d(512),)self.conv_res6 = torch.nn.Sequential(torch.nn.Conv2d(in_channels=512, out_channels=512, kernel_size=3, padding=1, stride=1, bias=False),torch.nn.BatchNorm2d(512),self.relu,torch.nn.Conv2d(in_channels=512, out_channels=512, kernel_size=3, padding=1, bias=False),torch.nn.BatchNorm2d(512),)def forward(self,x):x = self.block1(x)identity = self.downsample(x)x = self.conv_res(x)x += identityx = self.relu(x)x = self.conv_res2(x)x += identityx = self.relu(x)identity = self.downsample2(x)x = self.conv_res3(x)x += identityx = self.relu(x)x = self.conv_res4(x)x += identityx = self.relu(x)identity = self.downsample3(x)x = self.conv_res5(x)x += identityx = self.relu(x)x = self.conv_res6(x)x += identityx = self.relu(x)x = self.adavgpool(x)x = x.view(len(x), -1)x = self.output(x)return x

注意：此模型不是完整的ResNet网络，这里做了部分修改，以适应当前图片尺寸。

4.定义损失函数、优化器

from torch.optim import Adam
from torch.nn import functional as F# 初始化一个模型，输入图片通道数为1，输出特征为10
model = Model().to(device)
# 使用负对数似然损失函数
criterion = torch.nn.CrossEntropyLoss()
# 初始化Adam优化器，设定学习率为0.005
opt = Adam(model.parameters(), lr=0.001)

5.开始训练

# 进行9次迭代
for _ in range(49):# 遍历数据批次for n_, batch in enumerate(train_batch):# 将输入数据X调整形状并输入到模型X = batch[0].to(device)# y为真实标签y = batch[1].to(device)# 前向传播，获取模型输出sigma = model.forward(X)# 计算损失loss = criterion(sigma, y)# 计算预测的标签y_hat = torch.max(sigma, dim=1)[1]# 计算预测正确的数量correct_count = torch.sum(y_hat == y)# 计算准确率accuracy = correct_count / len(y) * 100# 反向传播，计算梯度loss.backward()# 更新模型参数opt.step()# 清除之前的梯度model.zero_grad()# 打印当前批次的损失和准确率print(n_, 'loss:', loss.item(), 'accuracy:', accuracy.item())

输出：

468 loss: 0.21156974136829376 accuracy: 91.66667175292969
468 loss: 0.24343211948871613 accuracy: 89.58333587646484
468 loss: 0.3186508119106293 accuracy: 87.5
468 loss: 0.16633149981498718 accuracy: 93.75
468 loss: 0.13033141195774078 accuracy: 93.75
468 loss: 0.09412961453199387 accuracy: 97.91667175292969
468 loss: 0.044871985912323 accuracy: 98.95833587646484
468 loss: 0.023767223581671715 accuracy: 100.0
468 loss: 0.09273606538772583 accuracy: 97.91667175292969
...

6.验证测试集

correct_count = 0
for batch in test_batch:test_X = batch[0].to(device)test_y = batch[1].to(device)sigma = model.forward(torch.tensor(test_X, dtype=torch.float32))y_hat = torch.max(sigma, dim=1)[1]correct_count += torch.sum(y_hat == test_y)accuracy = correct_count / 10000 * 100
print('accuracy:', accuracy.item())