- 🍨 本文为🔗365天深度学习训练营 中的学习记录博客
- 🍖 原作者:K同学啊 | 接辅导、项目定制
一、本周学习内容
1、了解如何设置动态学习率(重点)
2、保存训练过程中的最佳模型权重
3、调整代码使测试集accuracy到达86%。
二、前言
本次使用pytorch实现对运动鞋品牌的识别,标签分为两类:‘adidas’和’nike’。数据集下载地址:链接: link。
三、电脑环境
电脑系统:macOS Monterey 12.7.4
语言环境:Python 3.8.0
编译器:Jupyter Notebook
深度学习环境:torch:2.2.1 torchvision:0.17.1
四、前期准备
1、导入相关依赖项,设置CPU
import torch
import torch.nn as nn
import torchvision.transforms as transforms
import torchvision
from torchvision import transforms, datasets
import os,PIL,pathlib
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
device
2、加载数据
import os,PIL,random,pathlib
data_dir = './data/'
data_dir = pathlib.Path(data_dir)
data_paths = list(data_dir.glob('*'))
classeNames = [str(path).split("/")[1] for path in data_paths]
classeNames
输出为:
[‘test’, ‘train’]
3、数据可视化
import matplotlib.pyplot as plt
from PIL import Image
# 指定图像文件夹路径
image_folder = './data/train/nike/'
image_files = [f for f in os.listdir(image_folder) if f.endswith((".jpg", ".png", ".jpeg"))]
fig, axes = plt.subplots(3, 8, figsize=(16, 6))
for ax, img_file in zip(axes.flat, image_files):
img_path = os.path.join(image_folder, img_file)
img = Image.open(img_path)
ax.imshow(img)
ax.axis('off')
plt.tight_layout()
plt.show()
图形变换操作:
# 关于transforms.Compose的更多介绍可以参考:https://blog.csdn.net/qq_38251616/article/details/124878863
train_transforms = transforms.Compose([
transforms.Resize([224, 224]), # 将输入图片resize成统一尺寸
# transforms.RandomHorizontalFlip(), # 随机水平翻转
transforms.ToTensor(), # 将PIL Image或numpy.ndarray转换为tensor,并归一化到[0,1]之间
transforms.Normalize( # 标准化处理-->转换为标准正太分布(高斯分布),使模型更容易收敛
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]) # 其中 mean=[0.485,0.456,0.406]与std=[0.229,0.224,0.225] 从数据集中随机抽样计算得到的。
])
test_transform = transforms.Compose([
transforms.Resize([224, 224]), # 将输入图片resize成统一尺寸
transforms.ToTensor(), # 将PIL Image或numpy.ndarray转换为tensor,并归一化到[0,1]之间
transforms.Normalize( # 标准化处理-->转换为标准正太分布(高斯分布),使模型更容易收敛
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]) # 其中 mean=[0.485,0.456,0.406]与std=[0.229,0.224,0.225] 从数据集中随机抽样计算得到的。
])
train_dataset = datasets.ImageFolder("./data/train/",transform=train_transforms)
test_dataset = datasets.ImageFolder("./data/test/",transform=train_transforms)
将数据集类别映射为数字索引:
数据集有两个类别:adidas和nike。
train_dataset.class_to_idx
输出为:
{‘adidas’: 0, ‘nike’: 1}
4、划分数据集
batch_size = 32
train_dl = torch.utils.data.DataLoader(train_dataset,
batch_size=batch_size,
shuffle=True,
num_workers=1)
test_dl = torch.utils.data.DataLoader(test_dataset,
batch_size=batch_size,
shuffle=True,
num_workers=1)
查看数据格式:
for X, y in test_dl:
print("Shape of X [N, C, H, W]: ", X.shape)
print("Shape of y: ", y.shape, y.dtype)
break
输出为:
Shape of X [N, C, H, W]: torch.Size([32, 3, 224, 224])
Shape of y: torch.Size([32]) torch.int64
五、搭建简单的CNN网络
import torch.nn.functional as F
class Model(nn.Module):
def __init__(self):
super(Model, self).__init__()
self.conv1 = nn.Sequential(
nn.Conv2d(3, 12, kernel_size=5, padding=0),
nn.BatchNorm2d(12),
nn.ReLU()
)
self.conv2 = nn.Sequential(
nn.Conv2d(12, 24, kernel_size=5, padding=0),
nn.BatchNorm2d(24),
nn.ReLU()
)
self.conv3 = nn.Sequential(
nn.Conv2d(24, 24, kernel_size=5, padding=0),
nn.BatchNorm2d(24),
nn.ReLU()
)
self.conv4 = nn.Sequential(
nn.Conv2d(24, 48, kernel_size=5, padding=0),
nn.BatchNorm2d(48),
nn.ReLU()
)
self.conv5 = nn.Sequential(
nn.Conv2d(48, 48, kernel_size=5, padding=0),
nn.BatchNorm2d(48),
nn.ReLU()
)
self.conv6 = nn.Sequential(
nn.Conv2d(48, 96, kernel_size=5, padding=0),
nn.BatchNorm2d(96),
nn.ReLU()
)
self.pool = nn.Sequential(
nn.MaxPool2d(2)
)
self.dropout = nn.Sequential(
nn.Dropout(0.2)
)
self.fc1 = nn.Sequential(
nn.Linear(96*47*47, len(classeNames))
)
def forward(self, x):
batch_size = x.size(0)
x = self.conv1(x) # 224 -> 220
x = self.conv2(x) # 220 -> 216
x = self.conv3(x) # 216 -> 212
x = self.pool(x) # 212 -> 106
x = self.conv4(x) # 106 -> 102
x = self.conv5(x) # 102 -> 98
x = self.conv6(x) # 98 -> 94
x = self.pool(x) # 94 -> 47
x = self.dropout(x)
x = x.view(batch_size, -1)
x = self.fc1(x)
return x
device = "cuda" if torch.cuda.is_available() else "cpu"
print(f"Using {device} device")
model = Model().to(device)
model
六、训练模型
1、编写训练函数
def train(dataloader, model, loss_fn, optimizer):
size = len(dataloader.dataset)
num_batches = len(dataloader)
train_acc, train_loss = 0, 0
for X, y in dataloader:
X, y = X.to(device), y.to(device)
pred = model(X)
loss = loss_fn(pred, y)
# 反向传播
optimizer.zero_grad() # grad属性归零
loss.backward() # 反向传播
optimizer.step() # 每一步自动更新
train_acc += (pred.argmax(1) == y).type(torch.float).sum().item()
train_loss += loss.item()
train_acc /= size
train_loss /= num_batches
return train_acc, train_loss
2、编写测试函数
def test(dataloader, model, loss_fn):
size = len(dataloader.dataset)
num_batches = len(dataloader)
test_acc, test_loss = 0, 0
# 当不进行训练时,停止梯度更新,节省计算内存消耗
with torch.no_grad():
for imgs, target in dataloader:
imgs, target = imgs.to(device), target.to(device)
target_pred = model(imgs)
loss = loss_fn(target_pred, target)
test_acc += (target_pred.argmax(1) == target).type(torch.float).sum().item()
test_loss += loss.item()
test_acc /= size
test_loss /= num_batches
return test_acc, test_loss
3、设置动态学习率
def adjust_learning_rate(optimizer, epoch, start_lr):
# 每 2 个epoch衰减到原来的 0.92
lr = start_lr * (0.92 ** (epoch // 2))
# 遍历优化器中的所有参数组,并将每个参数组的学习率设置为新计算出的 lr。
for param_group in optimizer.param_groups:
param_group['lr'] = lr
learn_rate = 1e-4 # 初始学习率
optimizer = torch.optim.SGD(model.parameters(), lr=learn_rate)
✨调用官方动态学习率接口,与上面方法是等价的:
# # 调用官方动态学习率接口时使用
# lambda1 = lambda epoch: (0.92 ** (epoch // 2)
# optimizer = torch.optim.SGD(model.parameters(), lr=learn_rate)
# scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer, lr_lambda=lambda1) #选定调整方法
4、正式训练
loss_fn = nn.CrossEntropyLoss() # 创建损失函数
epochs = 40
train_loss = []
train_acc = []
test_loss = []
test_acc = []
for epoch in range(epochs):
# 更新学习率(使用自定义学习率时使用)
adjust_learning_rate(optimizer, epoch, learn_rate)
model.train()
epoch_train_acc, epoch_train_loss = train(train_dl, model, loss_fn, optimizer)
# scheduler.step() # 更新学习率(调用官方动态学习率接口时使用)
model.eval()
epoch_test_acc, epoch_test_loss = test(test_dl, model, loss_fn)
train_acc.append(epoch_train_acc)
train_loss.append(epoch_train_loss)
test_acc.append(epoch_test_acc)
test_loss.append(epoch_test_loss)
# 获取当前的学习率
lr = optimizer.state_dict()['param_groups'][0]['lr']
template = ('Epoch:{:2d}, Train_acc:{:.1f}%, Train_loss:{:.3f}, Test_acc:{:.1f}%, Test_loss:{:.3f}, Lr:{:.2E}')
print(template.format(epoch+1, epoch_train_acc*100, epoch_train_loss,
epoch_test_acc*100, epoch_test_loss, lr))
print('Done')
输出结果:
七、结果可视化
1、Loss与Accuracy图
import matplotlib.pyplot as plt
#隐藏警告
import warnings
warnings.filterwarnings("ignore") #忽略警告信息
plt.rcParams['font.sans-serif'] = ['SimHei'] # 用来正常显示中文标签
plt.rcParams['axes.unicode_minus'] = False # 用来正常显示负号
plt.rcParams['figure.dpi'] = 100 #分辨率
epochs_range = range(epochs)
plt.figure(figsize=(12, 3))
plt.subplot(1, 2, 1)
plt.plot(epochs_range, train_acc, label='Training Accuracy')
plt.plot(epochs_range, test_acc, label='Test Accuracy')
plt.legend(loc='lower right')
plt.title('Training and Validation Accuracy')
plt.subplot(1, 2, 2)
plt.plot(epochs_range, train_loss, label='Training Loss')
plt.plot(epochs_range, test_loss, label='Test Loss')
plt.legend(loc='upper right')
plt.title('Training and Validation Loss')
plt.show()
2. 指定图片进行预测
torch.squeeze()函数和torch.unsqueeze()函数在之前的文章中有介绍。
from PIL import Image
classes = list(train_dataset.class_to_idx)
def predict_one_image(image_path, model, transform, classes):
test_img = Image.open(image_path).convert('RGB')
# plt.imshow(test_img) # 展示预测的图片
test_img = transform(test_img)
img = test_img.to(device).unsqueeze(0)
model.eval()
output = model(img)
_,pred = torch.max(output,1)
pred_class = classes[pred]
print(f'预测结果是:{pred_class}')
从adidas文件夹随机选一张图片进行预测:
# 预测训练集中的某张照片
predict_one_image(image_path='./data/test/adidas/1.jpg',
model=model,
transform=train_transforms,
classes=classes)
输出:
预测结果是:adidas
八、保存并加载模型
# 模型保存
PATH = './model.pth' # 保存的参数文件名
torch.save(model.state_dict(), PATH)
# 将参数加载到model当中
model.load_state_dict(torch.load(PATH, map_location=device))
九、动态学习率
1. torch.optim.lr_scheduler.StepLR
等间隔动态调整方法,每经过step_size个epoch,做一次学习率decay,以gamma值为缩小倍数。
函数原型:
torch.optim.lr_scheduler.StepLR(optimizer, step_size, gamma=0.1, last_epoch=-1)
关键参数详解:
- optimizer(Optimizer):是之前定义好的需要优化的优化器的实例名
- step_size(int):是学习率衰减的周期,每经过每个epoch,做一次学习率decay
- gamma(float):学习率衰减的乘法因子。Default:0.1
用法示例:
optimizer = torch.optim.SGD(net.parameters(), lr=0.001 )
scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=5, gamma=0.1)
2. lr_scheduler.LambdaLR
根据自己定义的函数更新学习率。
函数原型:
torch.optim.lr_scheduler.LambdaLR(optimizer, lr_lambda, last_epoch=-1, verbose=False)
关键参数详解:
- optimizer(Optimizer):是之前定义好的需要优化的优化器的实例名
- lr_lambda(function):更新学习率的函数
用法示例:
lambda1 = lambda epoch: (0.92 ** (epoch // 2) # 第二组参数的调整方法
optimizer = torch.optim.SGD(model.parameters(), lr=learn_rate)
scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer, lr_lambda=lambda1) #选定调整方法
3.lr_scheduler.MultiStepLR
在特定的 epoch 中调整学习率
函数原型:
torch.optim.lr_scheduler.MultiStepLR(optimizer, milestones, gamma=0.1, last_epoch=-1, verbose=False)
关键参数详解:
- optimizer(Optimizer):是之前定义好的需要优化的优化器的实例名
- milestones(list):是一个关于epoch数值的list,表示在达到哪个epoch范围内开始变化,必须是升序排列
- gamma(float):学习率衰减的乘法因子。Default:0.1
用法示例:
optimizer = torch.optim.SGD(net.parameters(), lr=0.001 )
scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer,
milestones=[2,6,15], #调整学习率的epoc数
gamma=0.1)
更多的官方动态学习率设置方式可参考:https://pytorch.org/docs/stable/optim.html
👉调用官方接口示例:
model = [Parameter(torch.randn(2, 2, requires_grad=True))]
optimizer = SGD(model, 0.1)
scheduler = ExponentialLR(optimizer, gamma=0.9)
for epoch in range(20):
for input, target in dataset:
optimizer.zero_grad()
output = model(input)
loss = loss_fn(output, target)
loss.backward()
optimizer.step()
scheduler.step()
总结:
对比k导给出的示例代码,我没有跑出很好的效果。
- 我尝试增大 Dropout(0.2) 层的参数,其他网络结构不变 -> 效果不好;
- 尝试按比例增加输出通道数,其他保持不变 -> 效果不好;
- 尝试增加卷积层(Dropout层参数保持不变和增加两种情况) -> 效果不好。
综合其他博主的文章,我尝试了上面代码给出的网络结构,三个“卷积-BN-激活”之后跟一个池化,三个“卷积-BN-激活”之后跟一个池化,跑出来的结果完成了要求,测试集准确率达到了88%。
对于网络层的结构和通道数的设置,还是要多尝试几次,寻找效果较好的结构。