目录
往期回顾
模型训练数据处理
一、DataLoader
1.1torch.utils.data.DataLoader
功能:构建可迭代的数据装载器
常用参数
• dataset: Dataset类,决定数据从哪读取及如何读取
• batchsize : 批大小
• num_works: 是否多进程读取数据
• shuffle: 每个epoch是否乱序
• drop_last:当样本数不能被batchsize整除时,是否舍弃最后一批数据
# 构建MyDataset实例 用来获取图片路径以及图片类别,图片数量
train_data = RMBDataset(data_dir=train_dir, transform=train_transform)
valid_data = RMBDataset(data_dir=valid_dir, transform=valid_transform)
# 构建DataLoder
train_loader = DataLoader(dataset=train_data, batch_size=BATCH_SIZE, shuffle=True)
valid_loader = DataLoader(dataset=valid_data, batch_size=BATCH_SIZE)
当使用dataset获取到每张图片路径的时候可以使用DataLoader对数据图片进行打包封装。
1.2数据常见概念
Epoch: 所有训练样本都已输入到模型中,称为一个Epoch
Iteration:一批样本输入到模型中,称之为一个Iteration
Batchsize:批大小,决定一个Epoch有多少个Iteration
举例:样本总数:80, Batchsize:8 ;1 Epoch 为80/8=10 Iteration
样本总数:87, Batchsize:8
当drop_last = True时,1 Epoch = 10 Iteration ,因为舍弃了后面7个样本
当 drop_last = False时,1 Epoch = 11 Iteration,后面7个样本保留为一个Iteration
二、Dataset
2.1torch.utils.data.Dataset
功能:Dataset抽象类,所有自定义的
Dataset需要继承它,并且复写__getitem__()
getitem :接收一个索引,返回一个样本
# 构建MyDataset实例 用来获取图片路径以及图片类别,图片数量
train_data = RMBDataset(data_dir=train_dir, transform=train_transform)
valid_data = RMBDataset(data_dir=valid_dir, transform=valid_transform)
import numpy as np
import torch
import os
import random
from PIL import Image
from torch.utils.data import Dataset
random.seed(1)
rmb_label = {"1": 0, "100": 1}
class RMBDataset(Dataset):
def __init__(self, data_dir, transform=None):
"""
rmb面额分类任务的Dataset
:param data_dir: str, 数据集所在路径
:param transform: torch.transform,数据预处理
"""
self.label_name = {"1": 0, "100": 1}
self.data_info = self.get_img_info(data_dir) # data_info存储所有图片路径和标签,在DataLoader中通过index读取样本
self.transform = transform
def __getitem__(self, index):
path_img, label = self.data_info[index]
img = Image.open(path_img).convert('RGB') # 0~255
if self.transform is not None:
img = self.transform(img) # 在这里做transform,转为tensor等等
return img, label
def __len__(self):
return len(self.data_info)
@staticmethod
def get_img_info(data_dir):
data_info = list()
for root, dirs, _ in os.walk(data_dir):
# 遍历类别
for sub_dir in dirs:
img_names = os.listdir(os.path.join(root, sub_dir))
img_names = list(filter(lambda x: x.endswith('.jpg'), img_names))
# 遍历图片
for i in range(len(img_names)):
img_name = img_names[i]
path_img = os.path.join(root, sub_dir, img_name)
label = rmb_label[sub_dir]
data_info.append((path_img, int(label)))
return data_info
2.2代码展示
思考一下下面三个问题:
文件夹里面的数据读取处理如下:
2.2.1划分数据集
以人民币二分类为案例,需要将人民币数据集划分为训练集、验证集、测试集;代码如下:
# -*- coding: utf-8 -*- # 设置文件编码格式为 UTF-8
import os # 导入处理操作系统相关功能的模块
import random # 导入随机数相关的模块
import shutil # 导入文件和目录操作相关的模块
def makedir(new_dir): # 定义一个函数,用于创建目录
if not os.path.exists(new_dir): # 如果目录不存在
os.makedirs(new_dir) # 创建目录
if __name__ == '__main__': # 程序入口点
random.seed(1) # 设置随机数种子为1,保证随机性可重复
dataset_dir = os.path.join("..", "..", "data", "RMB_data") # 数据集目录路径
split_dir = os.path.join("..", "..", "data", "rmb_split") # 划分后数据集目录路径
train_dir = os.path.join(split_dir, "train") # 训练集目录路径
valid_dir = os.path.join(split_dir, "valid") # 验证集目录路径
test_dir = os.path.join(split_dir, "test") # 测试集目录路径
train_pct = 0.8 # 训练集占比
valid_pct = 0.1 # 验证集占比
test_pct = 0.1 # 测试集占比
for root, dirs, files in os.walk(dataset_dir): # 遍历数据集目录中的文件夹
for sub_dir in dirs: # 遍历每个文件夹
imgs = os.listdir(os.path.join(root, sub_dir)) # 获取文件夹中所有文件列表
imgs = list(filter(lambda x: x.endswith('.jpg'), imgs)) # 过滤出以'.jpg'结尾的文件
random.shuffle(imgs) # 随机打乱文件列表顺序
img_count = len(imgs) # 获取文件数量
train_point = int(img_count * train_pct) # 计算训练集划分点
valid_point = int(img_count * (train_pct + valid_pct)) # 计算验证集划分点
for i in range(img_count): # 遍历每个文件
if i < train_point: # 如果在训练集范围内
out_dir = os.path.join(train_dir, sub_dir) # 输出目录为训练集目录下的子文件夹
elif i < valid_point: # 如果在验证集范围内
out_dir = os.path.join(valid_dir, sub_dir) # 输出目录为验证集目录下的子文件夹
else: # 其余情况
out_dir = os.path.join(test_dir, sub_dir) # 输出目录为测试集目录下的子文件夹
makedir(out_dir) # 创建输出目录
target_path = os.path.join(out_dir, imgs[i]) # 目标路径为输出目录下的文件路径
src_path = os.path.join(dataset_dir, sub_dir, imgs[i]) # 源文件路径为数据集目录下的文件路径
shutil.copy(src_path, target_path) # 复制文件到目标路径
print('Class:{}, train:{}, valid:{}, test:{}'.format(sub_dir, train_point, valid_point-train_point,
img_count-valid_point)) # 输出每个类别的训练、验证和测试集数量
窗口输出结果:
Class:1, train:80, valid:10, test:10
Class:100, train:80, valid:10, test:10
2.2.2训练代码
# -*- coding: utf-8 -*-
import os
import random # 导入随机数生成库
import numpy as np # 导入numpy库
import torch # 导入pytorch库
import torch.nn as nn # 导入pytorch的神经网络模块
from torch.utils.data import DataLoader # 导入pytorch的数据加载模块
import torchvision.transforms as transforms # 导入图像转换模块
import torch.optim as optim # 导入优化器模块
from matplotlib import pyplot as plt # 导入matplotlib的绘图模块
from model.lenet import LeNet # 导入LeNet模型
from tools.my_dataset import RMBDataset # 导入自己构建的数据集
def set_seed(seed=1): # 设置随机种子
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
set_seed() # 设置随机种子
rmb_label = {"1": 0, "100": 1} # 构建分类标签字典
# 参数设置
MAX_EPOCH = 10
BATCH_SIZE = 16
LR = 0.01
log_interval = 10
val_interval = 1
# ============================ step 1/5 数据 ============================
split_dir = os.path.join("..", "..", "data", "rmb_split") # 拼接数据集路径
train_dir = os.path.join(split_dir, "train") # 训练数据路径
valid_dir = os.path.join(split_dir, "valid") # 验证数据路径
norm_mean = [0.485, 0.456, 0.406] # 图像规范化的均值
norm_std = [0.229, 0.224, 0.225] # 图像规范化的方差
train_transform = transforms.Compose([
transforms.Resize((32, 32)), # 将图片resize到32x32
transforms.RandomCrop(32, padding=4), # 进行随机裁剪
transforms.ToTensor(), # 转换为Tensor
transforms.Normalize(norm_mean, norm_std), # 规范化
])
valid_transform = transforms.Compose([
transforms.Resize((32, 32)), # 将图片resize到32x32
transforms.ToTensor(),
transforms.Normalize(norm_mean, norm_std), # 规范化
])
# 构建MyDataset实例
train_data = RMBDataset(data_dir=train_dir, transform=train_transform)
valid_data = RMBDataset(data_dir=valid_dir, transform=valid_transform)
# 构建DataLoder
train_loader = DataLoader(dataset=train_data, batch_size=BATCH_SIZE, shuffle=True) # 训练数据加载器
valid_loader = DataLoader(dataset=valid_data, batch_size=BATCH_SIZE) # 验证数据加载器
# ============================ step 2/5 模型 ============================
net = LeNet(classes=2) # 实例化LeNet模型
net.initialize_weights() # 初始化权重
# ============================ step 3/5 损失函数 ============================
criterion = nn.CrossEntropyLoss() # 选择损失函数,交叉熵损失
# ============================ step 4/5 优化器 ============================
optimizer = optim.SGD(net.parameters(), lr=LR, momentum=0.9) # 选择优化器,SGD随机梯度下降
scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=10, gamma=0.1) # 设置学习率下降策略
# ============================ step 5/5 训练 ============================
train_curve = list()
valid_curve = list()
for epoch in range(MAX_EPOCH):
loss_mean = 0.
correct = 0.
total = 0.
net.train() # 设置模型为训练模式
for i, data in enumerate(train_loader): # 遍历训练数据
# forward,前向传播
inputs, labels = data
outputs = net(inputs)
# backward,反向传播
optimizer.zero_grad() # 梯度置零
loss = criterion(outputs, labels) # 计算loss
loss.backward() # 反向传播计算梯度
# update weights,更新权重
optimizer.step()
# 统计分类情况
_, predicted = torch.max(outputs.data, 1)
total += labels.size(0)
correct += (predicted == labels).squeeze().sum().numpy()
# 打印训练信息
loss_mean += loss.item()
train_curve.append(loss.item())
if (i+1) % log_interval == 0:
loss_mean = loss_mean / log_interval
print("Training:Epoch[{:0>3}/{:0>3}] Iteration[{:0>3}/{:0>3}] Loss: {:.4f} Acc:{:.2%}".format(
epoch, MAX_EPOCH, i+1, len(train_loader), loss_mean, correct / total))
loss_mean = 0.
scheduler.step() # 更新学习率
# validate the model
if (epoch+1) % val_interval == 0:
correct_val = 0.
total_val = 0.
loss_val = 0.
net.eval() # 设置模型为评估模式
with torch.no_grad():
for j, data in enumerate(valid_loader):
inputs, labels = data
outputs = net(inputs)
loss = criterion(outputs, labels)
_, predicted = torch.max(outputs.data, 1)
total_val += labels.size(0)
correct_val += (predicted == labels).squeeze().sum().numpy()
loss_val += loss.item()
loss_val_epoch = loss_val / len(valid_loader)
valid_curve.append(loss_val_epoch)
print("Valid:\t Epoch[{:0>3}/{:0>3}] Iteration[{:0>3}/{:0>3}] Loss: {:.4f} Acc:{:.2%}".format(
epoch, MAX_EPOCH, j+1, len(valid_loader), loss_val_epoch, correct_val / total_val))
train_x = range(len(train_curve))
train_y = train_curve
train_iters = len(train_loader)
valid_x = np.arange(1, len(valid_curve)+1) * train_iters*val_interval # 转换x轴记录点
valid_y = valid_curve
plt.plot(train_x, train_y, label='Train') # 画训练loss图
plt.plot(valid_x, valid_y, label='Valid') # 画验证loss图
plt.legend(loc='upper right')
plt.ylabel('loss value')
plt.xlabel('Iteration')
plt.show()
# ============================ inference ============================
BASE_DIR = os.path.dirname(os.path.abspath(__file__))
test_dir = os.path.join(BASE_DIR, "test_data")
test_data = RMBDataset(data_dir=test_dir, transform=valid_transform)
valid_loader = DataLoader(dataset=test_data, batch_size=1)
for i, data in enumerate(valid_loader):
# forward
inputs, labels = data
outputs = net(inputs)
_, predicted = torch.max(outputs.data, 1)
rmb = 1 if predicted.numpy()[0] == 0 else 100
print("模型获得{}元".format(rmb))
三、transform
3.1 torchvision视觉工具包
torchvision.transforms : 常用的图像预处理方法:图像翻转、裁剪等
torchvision.datasets : 常用数据集的dataset实现,MNIST,CIFAR-10,ImageNet等
torchvision.model : 常用的模型预训练,AlexNet,VGG, ResNet,GoogLeNet等
torchvision.transforms : 常用的图像预处理方法
• 数据中心化
• 数据标准化
• 缩放
• 裁剪
• 旋转
• 翻转
• 填充
• 噪声添加
• 灰度变换
• 线性变换
• 仿射变换
• 亮度、饱和度及对比度变换
3.2 代码展示
Compose:将一系列的图像处理方法整合到一块包装。
split_dir = os.path.join("..", "..", "data", "rmb_split") # 从哪读取数据
train_dir = os.path.join(split_dir, "train")
valid_dir = os.path.join(split_dir, "valid")
norm_mean = [0.485, 0.456, 0.406]
norm_std = [0.229, 0.224, 0.225]
train_transform = transforms.Compose([
transforms.Resize((32, 32)), #设置图像大小
transforms.RandomCrop(32, padding=4), #随机裁剪
transforms.ToTensor(),
transforms.Normalize(norm_mean, norm_std), #将数据均值变为0,标准差变为1
])
valid_transform = transforms.Compose([
transforms.Resize((32, 32)),
transforms.ToTensor(),
transforms.Normalize(norm_mean, norm_std),
])
3.3 transforms.Normalize
功能:逐channel的对图像进行标准化,也就是output = (input - mean) / std,可以加快模型的收敛。
• mean:各通道的均值
• std:各通道的标准差
• inplace:是否原地操作
代码展示:
def normalize(tensor: Tensor, mean: List[float], std: List[float], inplace: bool = False) -> Tensor:
_assert_image_tensor(tensor)
if not tensor.is_floating_point():
raise TypeError(f"Input tensor should be a float tensor. Got {tensor.dtype}.")
if tensor.ndim < 3:
raise ValueError(
f"Expected tensor to be a tensor image of size (..., C, H, W). Got tensor.size() = {tensor.size()}"
)
if not inplace:
tensor = tensor.clone()
dtype = tensor.dtype
mean = torch.as_tensor(mean, dtype=dtype, device=tensor.device)
std = torch.as_tensor(std, dtype=dtype, device=tensor.device)
if (std == 0).any():
raise ValueError(f"std evaluated to zero after conversion to {dtype}, leading to division by zero.")
if mean.ndim == 1:
mean = mean.view(-1, 1, 1)
if std.ndim == 1:
std = std.view(-1, 1, 1)
return tensor.sub_(mean).div_(std) #减均值并且除以标准差
3.4 Normalize实验
为什么Normalize可以加快模型的收敛,以逻辑回归代码为例,此时bias = 1,均值接近0,当acc > 0.99停止训练,在380轮左右精确度达到99.5%
# -*- coding: utf-8 -*-
import torch
import torch.nn as nn
import matplotlib.pyplot as plt
import numpy as np
torch.manual_seed(10)
# ============================ step 1/5 生成数据 ============================
sample_nums = 100
mean_value = 1.7
bias = 1
n_data = torch.ones(sample_nums, 2)
x0 = torch.normal(mean_value * n_data, 1) + bias # 类别0 数据 shape=(100, 2)
y0 = torch.zeros(sample_nums) # 类别0 标签 shape=(100, 1)
x1 = torch.normal(-mean_value * n_data, 1) + bias # 类别1 数据 shape=(100, 2)
y1 = torch.ones(sample_nums) # 类别1 标签 shape=(100, 1)
train_x = torch.cat((x0, x1), 0) # 在0维进行拼接
train_y = torch.cat((y0, y1), 0)
# ============================ step 2/5 选择模型 ============================
class LR(nn.Module):
def __init__(self):
super(LR, self).__init__()
self.features = nn.Linear(2, 1)
self.sigmoid = nn.Sigmoid()
def forward(self, x):
x = self.features(x)
x = self.sigmoid(x)
return x
lr_net = LR() # 实例化逻辑回归模型
# ============================ step 3/5 选择损失函数 ============================
loss_fn = nn.BCELoss() # 二分类交叉熵损失函数
# ============================ step 4/5 选择优化器 ============================
lr = 0.01 # 学习率
optimizer = torch.optim.SGD(lr_net.parameters(), lr=lr, momentum=0.9)
# ============================ step 5/5 模型训练 ============================
for iteration in range(1000):
# 前向传播
y_pred = lr_net(train_x)
# 计算 loss
loss = loss_fn(y_pred.squeeze(), train_y)
# 反向传播
loss.backward()
# 更新参数
optimizer.step()
# 清空梯度
optimizer.zero_grad()
# 绘图
if iteration % 20 == 0:
mask = y_pred.ge(0.5).float().squeeze() # 以0.5为阈值进行分类
correct = (mask == train_y).sum() # 计算正确预测的样本个数
acc = correct.item() / train_y.size(0) # 计算分类准确率
plt.scatter(x0.data.numpy()[:, 0], x0.data.numpy()[:, 1], c='r', label='class 0')
plt.scatter(x1.data.numpy()[:, 0], x1.data.numpy()[:, 1], c='b', label='class 1')
w0, w1 = lr_net.features.weight[0]
w0, w1 = float(w0.item()), float(w1.item())
plot_b = float(lr_net.features.bias[0].item())
plot_x = np.arange(-6, 6, 0.1)
plot_y = (-w0 * plot_x - plot_b) / w1
plt.xlim(-5, 7)
plt.ylim(-7, 7)
plt.plot(plot_x, plot_y)
plt.text(-5, 5, 'Loss=%.4f' % loss.data.numpy(), fontdict={'size': 20, 'color': 'red'})
plt.title("Iteration: {}\nw0:{:.2f} w1:{:.2f} b: {:.2f} accuracy:{:.2%}".format(iteration, w0, w1, plot_b, acc))
plt.legend()
plt.show()
plt.pause(0.5)
if acc > 0.99:
break
当修改bias = 5,均值远离0,当acc > 0.99停止训练,在580轮左右精确度还未达到99%。