pytorch实现运动鞋分类

•   本文为365天深度学习训练营 中的学习记录博客
• 参考文章地址： 365天深度学习训练营-第P5周：运动鞋分类
• 作者：K同学啊

一、前期准备

1. 设置GPU

import torch
from torch import nn
import torchvision
from torchvision import transforms,datasets,models
import matplotlib.pyplot as plt
import os,PIL,pathlib

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
device

device(type='cuda')

2. 导入数据

data_dir = './data/'
data_dir = pathlib.Path(data_dir)

image_count = len(list(data_dir.glob('*/*/*.jpg')))
print("图片总数为：",image_count)

图片总数为： 578

classNames = [str(path).split('\\')[2] for path in data_dir.glob('train/*/')]
classNames

['adidas', 'nike']

roses= list(data_dir.glob('train/nike/*.jpg'))
PIL.Image.open(str(roses[0]))

 

3. 数据增强 解决过拟合

train_transforms = transforms.Compose([
        transforms.Resize([224, 224]),
       transforms.RandomRotation(45),#随机旋转，-45到45度之间随机选
#       transforms.CenterCrop(224),#从中心开始裁剪
        transforms.RandomHorizontalFlip(p=0.5),#随机水平翻转 选择一个概率概率
#         transforms.RandomVerticalFlip(p=0.5),#随机垂直翻转
#         transforms.ColorJitter(brightness=0.2, contrast=0.1, saturation=0.1, hue=0.1),#参数1为亮度，参数2为对比度，参数3为饱和度，参数4为色相
#         transforms.RandomGrayscale(p=0.025),#概率转换成灰度率，3通道就是R=G=B
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])#均值，标准差
    ])

test_transforms = transforms.Compose([
        transforms.Resize([224, 224]),
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
    ])

batch_size = 32

train_dataset = datasets.ImageFolder('./data/train/', transform = train_transforms)
test_dataset = datasets.ImageFolder('./data/test/', transform = test_transforms)


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)

classNames = train_dataset.classes

train_dataset.class_to_idx

{'adidas': 0, 'nike': 1}

4. 数据可视化

imgs, labels = next(iter(train_dl))
imgs.shape

import numpy as np

 # 指定图片大小，图像大小为20宽、5高的绘图(单位为英寸inch)
plt.figure(figsize=(20, 5)) 
for i, imgs in enumerate(imgs[:20]):
    npimg = imgs.numpy().transpose((1,2,0))
    npimg = npimg * np.array((0.229, 0.224, 0.225)) + np.array((0.485, 0.456, 0.406))
    npimg = npimg.clip(0, 1)
    # 将整个figure分成2行10列，绘制第i+1个子图。
    plt.subplot(2, 10, i+1)
    plt.imshow(npimg)
    plt.axis('off')

for X,y in test_dl:
    print('Shape of X [N, C, H, W]:', X.shape)
    print('Shape of y:', y.shape)
    break

Shape of X [N, C, H, W]: torch.Size([32, 3, 224, 224])
Shape of y: torch.Size([32])

二、构建CNN网络

2.1 搭建简单网络

搭建简单网络后发现由于数据量少导致过拟合，数据增强后最高准确率84%，说明模型不够好，选择改用Resnet18+迁移学习：

2.2 迁移学习

2.2.1 调用resnet18和预训练模型、冻结参数 

feature_extract = True

# 冻结参数
def set_parameter_requires_grad(model, feature_extracting):
    if feature_extracting:
        for param in model.parameters():
            param.requires_grad = False


# 修改输出层
def initialize_model(num_classes, feature_extract, use_pretrained=True):
    
    model_ft = models.resnet18(pretrained=use_pretrained)
    set_parameter_requires_grad(model_ft, feature_extract)
    
    num_ftrs = model_ft.fc.in_features
    model_ft.fc = nn.Linear(num_ftrs, num_classes)
                            
    input_size = 32

    return model_ft, input_size

model_ft, input_size = initialize_model(2, feature_extract, use_pretrained=True)
model_ft = model_ft.to(device)
model_ft

略

2.2.2 取出输出层参数 

取出输出层参数  后面用于训练更新 

# 设置训练哪些层
params_to_update = model_ft.parameters()
print("Params to learn:")
if feature_extract: # 自己只训练输出层
    params_to_update = []
    for name,param in model_ft.named_parameters():
        if param.requires_grad == True:
            params_to_update.append(param)
            print("\t",name)
else:
    for name,param in model_ft.named_parameters():
        if param.requires_grad == True:
            print("\t",name)

Params to learn:
	 fc.weight
	 fc.bias

三、训练模型

3.1 设置超参数

动态学习率 

# 优化器设置
optimizer = torch.optim.Adam(params_to_update, lr=1e-4)#要训练啥参数，你来定
scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=5, gamma=0.92)#学习率每7个epoch衰减成原来的1/10
loss_fn = nn.CrossEntropyLoss()

# def adjust_learning_rate(optimizer, epoch, start_lr):
#     # 每2个 epoch衰减到原来的0.98
#     lr = start_lr * (0.92 ** (epoch //2))
#     for param_group in optimizer.param_groups:
#         param_group['lr'] = lr
        
# optimizer = torch.optim.Adam(params_to_update,lr=1e-4)

3.2 编写训练函数

# 训练循环
def train(dataloader, model, loss_fn, optimizer):
    size = len(dataloader.dataset)  # 训练集的大小，一共900张图片
    num_batches = len(dataloader)   # 批次数目，29（900/32）

    train_loss, train_acc = 0, 0  # 初始化训练损失和正确率
    
    for X, y in dataloader:  # 获取图片及其标签
        X, y = X.to(device), y.to(device)
        
        # 计算预测误差
        pred = model(X)          # 网络输出
        loss = loss_fn(pred, y)  # 计算网络输出和真实值之间的差距，targets为真实值，计算二者差值即为损失
        
        # 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
        
        # 记录acc与loss
        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

3.3 编写测试函数

def test (dataloader, model, loss_fn):
    size        = len(dataloader.dataset)  # 测试集的大小，一共10000张图片
    num_batches = len(dataloader)          # 批次数目，8（255/32=8，向上取整）
    test_loss, test_acc = 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)
            
            # 计算loss
            target_pred = model(imgs)
            loss        = loss_fn(target_pred, target)
            
            test_loss += loss.item()
            test_acc  += (target_pred.argmax(1) == target).type(torch.float).sum().item()

    test_acc  /= size
    test_loss /= num_batches

    return test_acc, test_loss

3.4 正式训练

3.4.1 训练输出层

epochs     = 20
train_loss = []
train_acc  = []
test_loss  = []
test_acc   = []
best_acc = 0
filename='checkpoint.pth'


for epoch in range(epochs):
    model_ft.train()
    epoch_train_acc, epoch_train_loss = train(train_dl, model_ft, loss_fn, optimizer)
    
    scheduler.step()#学习率衰减
    
    model_ft.eval()
    epoch_test_acc, epoch_test_loss = test(test_dl, model_ft, loss_fn)
    
    # 保存最优模型
    if epoch_test_acc > best_acc:
        best_acc = epoch_test_acc
        state = {
            'state_dict': model_ft.state_dict(),#字典里key就是各层的名字，值就是训练好的权重
            'best_acc': best_acc,
            'optimizer' : optimizer.state_dict(),
        }
        torch.save(state, filename)
        
        
    train_acc.append(epoch_train_acc)
    train_loss.append(epoch_train_loss)
    test_acc.append(epoch_test_acc)
    test_loss.append(epoch_test_loss)
    
    template = ('Epoch:{:2d}, Train_acc:{:.1f}%, Train_loss:{:.3f}, Test_acc:{:.1f}%，Test_loss:{:.3f}')
    print(template.format(epoch+1, epoch_train_acc*100, epoch_train_loss, epoch_test_acc*100, epoch_test_loss))
print('Done')
print('best_acc：',best_acc)

Epoch:17, Train_acc:65.9%, Train_loss:0.630, Test_acc:67.1%，Test_loss:0.615
Epoch:18, Train_acc:66.1%, Train_loss:0.613, Test_acc:64.5%，Test_loss:0.599
Epoch:19, Train_acc:63.7%, Train_loss:0.636, Test_acc:65.8%，Test_loss:0.579
Epoch:20, Train_acc:66.3%, Train_loss:0.612, Test_acc:65.8%，Test_loss:0.583
Done
best_acc： 0.6593625498007968

3.4.2 训练所有层

for param in model_ft.parameters():
    param.requires_grad = True

# 再继续训练所有的参数，学习率调小一点
optimizer = torch.optim.Adam(model_ft.parameters(), lr=1e-4)
scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=5, gamma=0.92)

# 损失函数
criterion = nn.CrossEntropyLoss()

# 加载之前训练好的权重参数
checkpoint = torch.load(filename)
best_acc = checkpoint['best_acc']
model_ft.load_state_dict(checkpoint['state_dict'])

epochs     = 20
train_loss = []
train_acc  = []
test_loss  = []
test_acc   = []
best_acc = 0
filename='best_resnet18.pth'


for epoch in range(epochs):
    model_ft.train()
    epoch_train_acc, epoch_train_loss = train(train_dl, model_ft, loss_fn, optimizer)
    
    scheduler.step()#学习率衰减
    
    model_ft.eval()
    epoch_test_acc, epoch_test_loss = test(test_dl, model_ft, loss_fn)
    
    # 保存最优模型
    if epoch_test_acc > best_acc:
        best_acc = epoch_test_acc
        state = {
            'state_dict': model_ft.state_dict(),#字典里key就是各层的名字，值就是训练好的权重
            'best_acc': best_acc,
            'optimizer' : optimizer.state_dict(),
        }
        torch.save(state, filename)
        
        
    train_acc.append(epoch_train_acc)
    train_loss.append(epoch_train_loss)
    test_acc.append(epoch_test_acc)
    test_loss.append(epoch_test_loss)
    
    template = ('Epoch:{:2d}, Train_acc:{:.1f}%, Train_loss:{:.3f}, Test_acc:{:.1f}%，Test_loss:{:.3f}')
    print(template.format(epoch+1, epoch_train_acc*100, epoch_train_loss, epoch_test_acc*100, epoch_test_loss))
print('Done')
print('best_acc：',best_acc)

Epoch:18, Train_acc:99.8%, Train_loss:0.010, Test_acc:86.8%，Test_loss:0.398
Epoch:19, Train_acc:99.0%, Train_loss:0.031, Test_acc:93.4%，Test_loss:0.203
Epoch:20, Train_acc:99.2%, Train_loss:0.019, Test_acc:93.4%，Test_loss:0.184
Done
best_acc： 0.9342105263157895

四、结果可视化

加载训练好的模型

model_ft, input_size = initialize_model(2, feature_extract, use_pretrained=True)

# GPU模式
model_ft = model_ft.to(device)

# 保存文件的名字
filename='best_resnet18.pth'

# 加载模型
checkpoint = torch.load(filename)
best_acc = checkpoint['best_acc']
model_ft.load_state_dict(checkpoint['state_dict'])

结果可视化

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()

测试模型

train_on_gpu = True

# 得到一个batch的测试数据
imgs, labels = next(iter(train_dl))

# 进行预测
model_ft.eval()

if train_on_gpu:
    output = model_ft(imgs.cuda())
else:
    output = model_ft(imgs)


# 获得预测结果(概率最大的)
_, preds_tensor = torch.max(output, 1)

preds = np.squeeze(preds_tensor.numpy()) if not train_on_gpu else np.squeeze(preds_tensor.cpu().numpy())
preds

array([0, 1, 1, 0, 0, 1, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 1, 1, 1, 1,
       0, 0, 0, 0, 1, 0, 1, 0, 0, 1], dtype=int64)

import numpy as np

 # 指定图片大小，图像大小为20宽、5高的绘图(单位为英寸inch)
plt.figure(figsize=(20, 10)) 
for idx, imgs in enumerate(imgs[:10]):
    #ax = fig.add_subplot(rows, columns, idx+1, xticks=[], yticks=[])
    npimg = imgs.numpy().transpose((1,2,0))
    npimg = npimg * np.array((0.229, 0.224, 0.225)) + np.array((0.485, 0.456, 0.406))
    npimg = npimg.clip(0, 1)
    # 将整个figure分成2行10列，绘制第i+1个子图。
    ax = plt.subplot(2, 5, idx+1)
    ax.set_title("{} ({})".format(classNames[preds[idx]], classNames[labels[idx]]),
                 color=("green" if classNames[preds[idx]]==classNames[labels[idx]] else "red"))
    plt.imshow(npimg)
    plt.axis('off')