Pytorch从零开始实战——运动鞋识别
本系列来源于365天深度学习训练营
原作者K同学
文章目录
- Pytorch从零开始实战——运动鞋识别
- 环境准备
- 数据集
- 模型选择
- 数据可视化
- 模型预测
- 总结
环境准备
本文基于Jupyter notebook，使用Python3.8，Pytorch2.0.1+cu118，torchvision0.15.2，需读者自行配置好环境且有一些深度学习理论基础。本次实验的目的是了解如何设置动态学习率。
第一步，导入常用包。
import torch
import torch.nn as nn
import matplotlib.pyplot as plt
import torchvision
import torchvision.transforms as transforms
import torchvision.datasets as datasets
import torch.nn.functional as F
import random
from time import time
import numpy as np
import pandas as pd
import datetime
import gc
import os
os.environ['KMP_DUPLICATE_LIB_OK']='True' # 用于避免jupyter环境突然关闭
torch.backends.cudnn.benchmark=True # 用于加速GPU运算的代码
设置随机数种子，428不好用，这次设置为55
torch.manual_seed(55)
torch.cuda.manual_seed(55)
torch.cuda.manual_seed_all(55)
random.seed(55)
np.random.seed(55)
创建设备对象，检测设备
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
device
数据集
本次实验是对运动鞋图片进行分类任务，共579张图片，是一个二分类任务，标签为adidas、nike，两种类别的图片分别存放在不同的文件夹中。
展示图片函数
def plotsample(data):
fig, axs = plt.subplots(1, 5, figsize=(10, 10)) #建立子图
for i in range(5):
num = random.randint(0, len(data) - 1) #首先选取随机数，随机选取五次
#抽取数据中对应的图像对象，make_grid函数可将任意格式的图像的通道数升为3，而不改变图像原始的数据
#而展示图像用的imshow函数最常见的输入格式也是3通道
npimg = torchvision.utils.make_grid(data[num][0]).numpy()
nplabel = data[num][1] #提取标签
#将图像由(3, weight, height)转化为(weight, height, 3)，并放入imshow函数中读取
axs[i].imshow(np.transpose(npimg, (1, 2, 0)))
axs[i].set_title(nplabel) #给每个子图加上标签
axs[i].axis("off") #消除每个子图的坐标轴
查看classNames
import pathlib
data_dir = './data/snk/train'
data_dir = pathlib.Path(data_dir) # 转成pathlib.Path对象
data_paths = list(data_dir.glob('*')) # [PosixPath('data/snk/train/adidas'), PosixPath('data/snk/train/nike')]
classNames = [str(path).split("/")[3] for path in data_paths]
classNames # 二分类问题 ['adidas', 'nike']
使用transforms来预处理原始数据，统一尺寸、转换为张量、标准化
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]) # 标准化
])
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]) # 标准化
])
# 根据文件名设置标签
train_dataset = datasets.ImageFolder("./data/snk/train/", transform=train_transforms)
test_dataset = datasets.ImageFolder("./data/snk/test/", transform=train_transforms)
随机查看5张图片
plotsample(train_dataset)
使用DataLoader划分数据集，batch_size = 32
batch_size = 32
train_dl = torch.utils.data.DataLoader(train_dataset,
batch_size=batch_size,
shuffle=True,
)
test_dl = torch.utils.data.DataLoader(test_dataset,
batch_size=batch_size,
shuffle=True,
)
len(train_dl.dataset), len(test_dl.dataset) # 503 76
模型选择
本次还是选择简单的卷积神经网络，这次写法使用Sequential，表示这一块是一个单独的模块。
class Model(nn.Module):
def __init__(self):
super().__init__()
self.conv1 = nn.Sequential(
nn.Conv2d(3, 12, kernel_size=5), # 220
nn.BatchNorm2d(12),
nn.ReLU()
)
self.conv2 = nn.Sequential(
nn.Conv2d(12, 12, kernel_size=5), # 216
nn.BatchNorm2d(12),
nn.ReLU()
)
self.pool3 = nn.Sequential(
nn.MaxPool2d(2) # 108
)
self.conv4 = nn.Sequential(
nn.Conv2d(12, 24, kernel_size=5), # 104
nn.BatchNorm2d(24),
nn.ReLU()
)
self.conv5 = nn.Sequential(
nn.Conv2d(24, 24, kernel_size=5), # 100
nn.BatchNorm2d(24),
nn.ReLU()
)
self.pool6 = nn.Sequential(
nn.MaxPool2d(2)
)
self.dropout = nn.Sequential(
nn.Dropout(0.2)
)
self.fc = nn.Sequential(
nn.Linear(50 * 50 * 24, len(classNames))
)
def forward(self, x):
x = self.conv1(x)
x = self.conv2(x)
x = self.pool3(x)
x = self.conv4(x)
x = self.conv5(x)
x = self.pool6(x)
x = self.dropout(x)
x = x.view(-1, 50 * 50 * 24)
x = self.fc(x)
return x
模型初始化
from torchsummary import summary
# 将模型转移到GPU中
model = Model().to(device)
summary(model, input_size=(3, 224, 224))
定义训练函数
def train(dataloader, model, loss_fn, opt):
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)
opt.zero_grad()
loss.backward()
opt.step()
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
定义测试函数
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 X, y in dataloader:
X, y = X.to(device), y.to(device)
pred = model(X)
loss = loss_fn(pred, y)
test_acc += (pred.argmax(1) == y).type(torch.float).sum().item()
test_loss += loss.item()
test_acc /= size
test_loss /= num_batches
return test_acc, test_loss
定义一些超参数
loss_fn = nn.CrossEntropyLoss()
learn_rate = 0.0001
opt = torch.optim.SGD(model.parameters(), lr=learn_rate)
定义学习率衰减函数，大概意思是随着epoch的增加，学习率会持续变小，使得模型更容易收敛
def adjust_rate(opt, epoch, start_lr):
lr = start_lr * (0.92 ** (epoch // 2))
for param_group in opt.param_groups:
param_group['lr'] = lr
开始训练
import time
epochs = 30
train_loss = []
train_acc = []
test_loss = []
test_acc = []
T1 = time.time()
best_acc = 0
PATH = './my_model.pth'
for epoch in range(epochs):
adjust_rate(opt, epoch, learn_rate)
model.train()
epoch_train_acc, epoch_train_loss = train(train_dl, model, loss_fn, opt)
model.eval() # 确保模型不会进行训练操作
epoch_test_acc, epoch_test_loss = test(test_dl, model, loss_fn)
if epoch_test_acc > best_acc:
best_acc = epoch_test_acc
torch.save(model.state_dict(), PATH)
print("model save")
train_acc.append(epoch_train_acc)
train_loss.append(epoch_train_loss)
test_acc.append(epoch_test_acc)
test_loss.append(epoch_test_loss)
print("epoch:%d, train_acc:%.1f%%, train_loss:%.3f, test_acc:%.1f%%, test_loss:%.3f"
% (epoch + 1, epoch_train_acc * 100, epoch_train_loss, epoch_test_acc * 100, epoch_test_loss))
print("Done")
T2 = time.time()
print('程序运行时间:%s毫秒' % ((T2 - T1)*1000))
但是效果好像不是很好，模型训练的时候卡在某个极小值不动了
经过实验，将学习率改为0.001，效果是最好的。
import time
epochs = 30
train_loss = []
train_acc = []
test_loss = []
test_acc = []
T1 = time.time()
best_acc = 0
PATH = './my_model.pth'
for epoch in range(epochs):
model.train()
epoch_train_acc, epoch_train_loss = train(train_dl, model, loss_fn, opt)
model.eval() # 确保模型不会进行训练操作
epoch_test_acc, epoch_test_loss = test(test_dl, model, loss_fn)
if epoch_test_acc > best_acc:
best_acc = epoch_test_acc
torch.save(model.state_dict(), PATH)
print("model save")
train_acc.append(epoch_train_acc)
train_loss.append(epoch_train_loss)
test_acc.append(epoch_test_acc)
test_loss.append(epoch_test_loss)
print("epoch:%d, train_acc:%.1f%%, train_loss:%.3f, test_acc:%.1f%%, test_loss:%.3f"
% (epoch + 1, epoch_train_acc * 100, epoch_train_loss, epoch_test_acc * 100, epoch_test_loss))
print("Done")
T2 = time.time()
print('程序运行时间:%s毫秒' % ((T2 - T1)*1000))
在训练集上已经达到百分百准确率了，在测试集上的表现也很好。
数据可视化
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()
模型预测
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) # 增加维度
model.eval()
output = model(img)
_,pred = torch.max(output,1)
pred_class = classes[pred]
print(f'预测结果是：{pred_class}')
使用2.jpg开始预测
predict_one_image(image_path='./data/snk/test/adidas/2.jpg',
model=model,
transform=train_transforms,
classes=classes)
预测结果是：adidas
总结
学习率衰减是一个很有用的东西，但有的时候，使用学习率衰减好像还不如不使用学习率衰减，感觉容易提前收敛。
