Pytorch实现图像分类-水果数据集分类--深度学习大作业 - ToB企服应用市场:ToB评测及商务社交产业平台

import torch
import torch.nn as nn
import torchvision
import torchvision.transforms as transforms
from torch.utils.data import DataLoader
import torch.nn.functional as F
from torch import nn, optim
# 数据预处理
image_size = (224, 224)
data_transforms = transforms.Compose([
transforms.RandomHorizontalFlip(),
transforms.Resize(image_size),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
])
#导入数据集
import torchvision.datasets as datasets
train_data=datasets.ImageFolder (root='fruits-360-original-size/fruits-360-original-size/Training',transform=data_transforms)
test_data=datasets.ImageFolder (root='fruits-360-original-size/fruits-360-original-size/Test',transform=data_transforms)
# print(train_data.classes)
# print('..............')
# print(test_data.classes)
#DataLoader
batchsize=10#每个批次（batch）中包含的样本数量
train_loader = DataLoader(train_data, batch_size=batchsize, shuffle=True, num_workers=1)
test_loader = DataLoader(test_data, batch_size=batchsize, shuffle=False, num_workers=1) # 测试时不需打乱数据
#创建模型
class AlexNet(nn.Module):
def __init__(self):
super(AlexNet, self).__init__()
self.conv = nn.Sequential(
nn.Conv2d(3, 96, 5, 1, 2),#输入通道数，输出通道数，卷积核大小，步长，填充（！！！rgb图像所以是三个通道，开始没注意以为灰度图像）
nn.ReLU(),
nn.MaxPool2d(3, 2),
nn.Conv2d(96, 256, 5, 1, 2),
nn.ReLU(),
nn.MaxPool2d(3, 2),
nn.Conv2d(256, 384, 3, 1, 1),
nn.ReLU(),
nn.Conv2d(384, 384, 3, 1, 1),
nn.ReLU(),
nn.Conv2d(384, 256, 3, 1, 1),
nn.ReLU(),
nn.MaxPool2d(3, 2)
)
# 计算全连接层输入大小
self.fc_input_size = self._get_fc_input_size()
self.fc = nn.Sequential(
nn.Linear(self.fc_input_size, 4096),
nn.ReLU(),
nn.Dropout(0.5),#随机丢弃
nn.Linear(4096, 4096),
nn.ReLU(),
nn.Dropout(0.5),
nn.Linear(4096, 2)#修改为2因为输出只有两个类
)
def forward(self, img):
# 通过卷积层前向传播，img是输入图像张量
feature = self.conv(img)
feature = feature.view(img.size(0), -1)#展平
# 通过全连接层（fc）进行前向传播，得到最终的输出
output = self.fc(feature)
return output
#动态计算全连接层（FC层）所需要的输入尺寸
def _get_fc_input_size(self):
# 创建一个与训练/测试时相同尺寸和通道数的随机张量，用于通过卷积层
x = torch.randn(1, 3, image_size[0], image_size[1])# 其中3对应RGB图像的通道数，image_size是从外部传入的图像预处理后的尺寸，默认为(224, 224)
x = self.conv(x)
return x.view(-1).size(0)#展平后的向量长度
# 实例化模型、损失函数和优化器
model = AlexNet().to(device="cuda" if torch.cuda.is_available() else "cpu")
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(), lr=0.001, momentum=0.9)
# 训练函数
def train(model, device, train_loader, optimizer, criterion):
model.train()
# 遍历训练数据加载器中的每个批次
for batch_idx, (data, target) in enumerate(train_loader):
data, target = data.to(device), target.to(device)#这样更快
# 梯度清零，防止梯度累积
optimizer.zero_grad()
output = model(data)#预测输出
loss = criterion(output, target)
loss.backward()
optimizer.step()
# 每10个batch打印一次训练信息
if (batch_idx + 1) % 10 == 0 or (batch_idx + 1) == len(train_loader):
print(f'训练轮次: {epoch + 1}/{num_epochs} 损失: {loss.item():.6f}')
# 测试函数
def test(model, device, test_loader):
model.eval()
test_loss = 0
correct = 0
#避免在测试过程中计算和存储梯度，节省内存并加速计算
with torch.no_grad():
for data, target in test_loader:
data, target = data.to(device), target.to(device)
output = model(data)
test_loss += criterion(output, target).item() # 累加批次损失
pred = output.argmax(dim=1, keepdim=True) # 获取预测概率最大的类别索引
correct += pred.eq(target.view_as(pred)).sum().item()#累加预测正确的数量
test_loss /= len(test_loader.dataset) # 平均损失
# 打印测试结果，包括平均损失、正确预测的总数、总样本数以及准确率
print(f'\n测试集: 平均损失: {test_loss:.6f}, 正确: {correct}/{len(test_loader.dataset)} ({100. * correct / len(test_loader.dataset):.2f}%)\n')
# 主训练循环
num_epochs = 10 # 设置训练轮数
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
for epoch in range(num_epochs):
print(f'第{epoch + 1}轮训练开始')
train(model, device, train_loader, optimizer, criterion)
test(model, device, test_loader)

复制代码