第五周--P5周：运动鞋识别-CSDN博客

本文链接：https://blog.csdn.net/weixin_46559155/article/details/134253004

>- **🍨 本文为[🔗365天深度学习训练营](https://mp.weixin.qq.com/s/Nb93582M_5usednAKp_Jtw) 中的学习记录博客**
>- **🍦 参考文章：[Pytorch实战 | 第P5周：运动鞋识别](https://www.heywhale.com/mw/project/6352467ca42e79f98f6bbf13)**
>- **🍖 原作者：[K同学啊 | 接辅导、项目定制](https://mtyjkh.blog.csdn.net/)**
>- **🚀 文章来源：[K同学的学习圈子](https://www.yuque.com/mingtian-fkmxf/zxwb45)**

一、前期准备

1. 设置GPU

import torch 
import torchvision
from torchvision import transforms,datasets
import torchvision.transforms as transforms
import torch.nn as nn

import os,PIL,pathlib

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

device

device(type='cuda')

2.导入数据

#查看类别
data_path   = r"F:\P5_data\test"
data_path   = pathlib.Path(data_path)
class_path  = list(data_path.glob("*"))
class_names = [str(i).split("\\")[-1] for i in class_path]
class_names

['adidas', 'nike']

#查看图片
import matplotlib.pyplot as plt
from PIL import Image

fig,axes     = plt.subplots(3, 8, figsize=(16, 6))
image_folder = r"F:\P5_data\train\adidas"
image_files  = [f for f in os.listdir(image_folder) if f.endswith((".jpg",".png","jepg"))]

for ax,file in zip(axes.flat,image_files):
    img_path = os.path.join(image_folder, file)
    img      = Image.open(img_path)
    ax.imshow(img)
    ax.axis('off')

import torchvision.transforms as transforms
train_transforms = transforms.Compose([
    transforms.Resize([224,224]),
    transforms.ToTensor(),
    transforms.Normalize(
    mean=[0.485,0.456,0.406],
    std=[0.229,0.224,0.225])
])

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

train_dataset = datasets.ImageFolder(r"F:\P5_data\train",transform=train_transforms)
test_dataset  = datasets.ImageFolder(r"F:\P5_data\test",transform=test_transforms)

train_dataset.class_to_idx
test_dataset.class_to_idx

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

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)

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

二、构建简单的CNN模型

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),#12*220*220
            nn.BatchNorm2d(12),
            nn.ReLU())
        self.conv2 = nn.Sequential(
            nn.Conv2d(12, 12, kernel_size=5, padding=0),#12*216*216
            nn.BatchNorm2d(12),
            nn.ReLU())
        self.pool1 = nn.Sequential(
            nn.MaxPool2d(2)) #12*108*108
        self.conv4 = nn.Sequential(
            nn.Conv2d(12, 24, kernel_size=5, padding=0),#24*104*104
            nn.BatchNorm2d(24),
            nn.ReLU())
        self.conv5 = nn.Sequential(
            nn.Conv2d(24, 24, kernel_size=5, padding=0),#24*100*100
            nn.BatchNorm2d(24),
            nn.ReLU())
        self.pool2 = nn.Sequential(
            nn.MaxPool2d(2)) #24*50*50
        self.conv7 = nn.Sequential(
            nn.Conv2d(24, 48, kernel_size=5, padding=0),#48*46*46
            nn.BatchNorm2d(48),
            nn.ReLU())
        self.conv8 = nn.Sequential(
            nn.Conv2d(48, 48, kernel_size=5, padding=0),#48*42*42
            nn.BatchNorm2d(48),
            nn.ReLU())
        self.pool3 = nn.Sequential(
            nn.MaxPool2d(2)) #48*21*21
        
        self.dropout =nn.Sequential(
            nn.Dropout(0.2))
        
        self.fc = nn.Sequential(
            nn.Linear(48*21*21,len(class_names)))
        
    def forward(self, x):
        
        batch_size = x.size(0)
        x = self.conv1(x) #卷积-BN-激活
        x = self.conv2(x) #卷积-BN-激活
        x = self.pool1(x) #池化层
        x = self.conv4(x)
        x = self.conv5(x)
        x = self.pool2(x)
        x = self.conv7(x)
        x = self.conv8(x)
        x = self.pool3(x)
        x = self.dropout(x)
        x = x.view(batch_size,-1) 
        x = self.fc(x) 
        
        return x
    
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print ("Using {} device:".format(device))

model = Model().to(device)
model

Using cuda device:

Out[8]:

Model(
  (conv1): Sequential(
    (0): Conv2d(3, 12, kernel_size=(5, 5), stride=(1, 1))
    (1): BatchNorm2d(12, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
  )
  (conv2): Sequential(
    (0): Conv2d(12, 12, kernel_size=(5, 5), stride=(1, 1))
    (1): BatchNorm2d(12, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
  )
  (pool1): Sequential(
    (0): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (conv4): Sequential(
    (0): Conv2d(12, 24, kernel_size=(5, 5), stride=(1, 1))
    (1): BatchNorm2d(24, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
  )
  (conv5): Sequential(
    (0): Conv2d(24, 24, kernel_size=(5, 5), stride=(1, 1))
    (1): BatchNorm2d(24, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
  )
  (pool2): Sequential(
    (0): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (conv7): Sequential(
    (0): Conv2d(24, 48, kernel_size=(5, 5), stride=(1, 1))
    (1): BatchNorm2d(48, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
  )
  (conv8): Sequential(
    (0): Conv2d(48, 48, kernel_size=(5, 5), stride=(1, 1))
    (1): BatchNorm2d(48, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU()
  )
  (pool3): Sequential(
    (0): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (dropout): Sequential(
    (0): Dropout(p=0.2, inplace=False)
  )
  (fc): Sequential(
    (0): Linear(in_features=21168, out_features=2, bias=True)
  )
)

三、训练模型

##编写训练函数

def train(dataloader, model, loss_fn, optimizer):
    size = len(dataloader.dataset) #训练集大小
    num_batches = len(dataloader) #batch个数
    
    
    train_loss, train_acc =0, 0 #初始化训练损失和正确率
    
    for X,y in dataloader:
        X,y = X.to(device),y.to(device)
        
        #计算预测误差
        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
        
        #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

##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
    
    with torch.no_grad():
        for imgs,target in dataloader:
            imgs, target =imgs.to(device),target.to(device)
            
            #计算loss和acc
            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

#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0.98
    lr = start_lr * (0.98**(epoch // 2))
    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)

#正式训练
loss_fn = nn.CrossEntropyLoss() #创建损失函数
epochs  = 50

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')

Epoch: 1, Train_acc:56.4%, Train_loss:0.722, Test_acc:50.0%, Test_loss:0.698, Lr:1.00E-04
Epoch: 2, Train_acc:58.4%, Train_loss:0.678, Test_acc:65.8%, Test_loss:0.657, Lr:1.00E-04
Epoch: 3, Train_acc:61.8%, Train_loss:0.669, Test_acc:73.7%, Test_loss:0.633, Lr:9.80E-05
Epoch: 4, Train_acc:65.3%, Train_loss:0.621, Test_acc:72.4%, Test_loss:0.577, Lr:9.80E-05
Epoch: 5, Train_acc:67.7%, Train_loss:0.607, Test_acc:73.7%, Test_loss:0.566, Lr:9.60E-05
Epoch: 6, Train_acc:67.3%, Train_loss:0.590, Test_acc:76.3%, Test_loss:0.571, Lr:9.60E-05
Epoch: 7, Train_acc:69.1%, Train_loss:0.578, Test_acc:72.4%, Test_loss:0.566, Lr:9.41E-05
Epoch: 8, Train_acc:71.7%, Train_loss:0.561, Test_acc:77.6%, Test_loss:0.552, Lr:9.41E-05
Epoch: 9, Train_acc:69.1%, Train_loss:0.543, Test_acc:82.9%, Test_loss:0.551, Lr:9.22E-05
Epoch:10, Train_acc:71.1%, Train_loss:0.551, Test_acc:81.6%, Test_loss:0.525, Lr:9.22E-05
Epoch:11, Train_acc:74.1%, Train_loss:0.509, Test_acc:81.6%, Test_loss:0.524, Lr:9.04E-05
Epoch:12, Train_acc:73.7%, Train_loss:0.517, Test_acc:81.6%, Test_loss:0.583, Lr:9.04E-05
Epoch:13, Train_acc:73.3%, Train_loss:0.504, Test_acc:78.9%, Test_loss:0.524, Lr:8.86E-05
Epoch:14, Train_acc:78.1%, Train_loss:0.498, Test_acc:84.2%, Test_loss:0.536, Lr:8.86E-05
Epoch:15, Train_acc:76.1%, Train_loss:0.493, Test_acc:84.2%, Test_loss:0.496, Lr:8.68E-05
Epoch:16, Train_acc:80.7%, Train_loss:0.468, Test_acc:84.2%, Test_loss:0.497, Lr:8.68E-05
Epoch:17, Train_acc:77.9%, Train_loss:0.469, Test_acc:85.5%, Test_loss:0.514, Lr:8.51E-05
Epoch:18, Train_acc:80.7%, Train_loss:0.463, Test_acc:85.5%, Test_loss:0.484, Lr:8.51E-05
Epoch:19, Train_acc:80.5%, Train_loss:0.454, Test_acc:85.5%, Test_loss:0.509, Lr:8.34E-05
Epoch:20, Train_acc:80.1%, Train_loss:0.452, Test_acc:76.3%, Test_loss:0.499, Lr:8.34E-05
Epoch:21, Train_acc:82.5%, Train_loss:0.441, Test_acc:84.2%, Test_loss:0.506, Lr:8.17E-05
Epoch:22, Train_acc:84.1%, Train_loss:0.420, Test_acc:81.6%, Test_loss:0.484, Lr:8.17E-05
Epoch:23, Train_acc:84.5%, Train_loss:0.421, Test_acc:85.5%, Test_loss:0.537, Lr:8.01E-05
Epoch:24, Train_acc:85.3%, Train_loss:0.412, Test_acc:88.2%, Test_loss:0.460, Lr:8.01E-05
Epoch:25, Train_acc:85.9%, Train_loss:0.410, Test_acc:86.8%, Test_loss:0.488, Lr:7.85E-05
Epoch:26, Train_acc:86.7%, Train_loss:0.392, Test_acc:80.3%, Test_loss:0.516, Lr:7.85E-05
Epoch:27, Train_acc:86.5%, Train_loss:0.395, Test_acc:82.9%, Test_loss:0.456, Lr:7.69E-05
Epoch:28, Train_acc:87.1%, Train_loss:0.396, Test_acc:88.2%, Test_loss:0.459, Lr:7.69E-05
Epoch:29, Train_acc:88.4%, Train_loss:0.387, Test_acc:86.8%, Test_loss:0.440, Lr:7.54E-05
Epoch:30, Train_acc:88.8%, Train_loss:0.370, Test_acc:84.2%, Test_loss:0.466, Lr:7.54E-05
Epoch:31, Train_acc:87.8%, Train_loss:0.379, Test_acc:88.2%, Test_loss:0.421, Lr:7.39E-05
Epoch:32, Train_acc:88.6%, Train_loss:0.363, Test_acc:86.8%, Test_loss:0.439, Lr:7.39E-05
Epoch:33, Train_acc:89.4%, Train_loss:0.360, Test_acc:73.7%, Test_loss:0.439, Lr:7.24E-05
Epoch:34, Train_acc:88.6%, Train_loss:0.359, Test_acc:85.5%, Test_loss:0.462, Lr:7.24E-05
Epoch:35, Train_acc:90.2%, Train_loss:0.350, Test_acc:86.8%, Test_loss:0.464, Lr:7.09E-05
Epoch:36, Train_acc:87.8%, Train_loss:0.352, Test_acc:88.2%, Test_loss:0.465, Lr:7.09E-05
Epoch:37, Train_acc:91.2%, Train_loss:0.333, Test_acc:86.8%, Test_loss:0.464, Lr:6.95E-05
Epoch:38, Train_acc:88.4%, Train_loss:0.353, Test_acc:86.8%, Test_loss:0.442, Lr:6.95E-05
Epoch:39, Train_acc:91.4%, Train_loss:0.335, Test_acc:88.2%, Test_loss:0.466, Lr:6.81E-05
Epoch:40, Train_acc:89.4%, Train_loss:0.336, Test_acc:88.2%, Test_loss:0.448, Lr:6.81E-05
Epoch:41, Train_acc:90.0%, Train_loss:0.339, Test_acc:89.5%, Test_loss:0.457, Lr:6.68E-05
Epoch:42, Train_acc:91.8%, Train_loss:0.328, Test_acc:88.2%, Test_loss:0.453, Lr:6.68E-05
Epoch:43, Train_acc:91.4%, Train_loss:0.315, Test_acc:89.5%, Test_loss:0.490, Lr:6.54E-05
Epoch:44, Train_acc:90.6%, Train_loss:0.323, Test_acc:86.8%, Test_loss:0.437, Lr:6.54E-05
Epoch:45, Train_acc:92.0%, Train_loss:0.320, Test_acc:88.2%, Test_loss:0.444, Lr:6.41E-05
Epoch:46, Train_acc:93.0%, Train_loss:0.312, Test_acc:88.2%, Test_loss:0.455, Lr:6.41E-05
Epoch:47, Train_acc:92.6%, Train_loss:0.314, Test_acc:89.5%, Test_loss:0.472, Lr:6.28E-05
Epoch:48, Train_acc:92.0%, Train_loss:0.305, Test_acc:86.8%, Test_loss:0.412, Lr:6.28E-05
Epoch:49, Train_acc:92.8%, Train_loss:0.303, Test_acc:90.8%, Test_loss:0.429, Lr:6.16E-05
Epoch:50, Train_acc:92.4%, Train_loss:0.308, Test_acc:90.8%, Test_loss:0.419, Lr:6.16E-05
Done

四、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()