Hands-on learning deep learning pytorch study notes
We will use the knowledge learned in the previous sections to participate in the Kaggle competition, which solves the CIFAR-10 image classification problem. The competition website is https://www.kaggle.com/c/cifar-10
# 本节的网络需要较长的训练时间
# 可以在Kaggle访问：
# https://www.kaggle.com/boyuai/boyu-d2l-image-classification-cifar-10
import numpy as np
import torch
import torch.nn as nn
import torch.optim as optim
import torchvision
import torchvision.transforms as transforms
import torch.nn.functional as F
import torch.optim as optim
from torchvision import datasets, transforms
import os
import time
Acquire and organize data sets
The competition data is divided into training set and test set. The training set contains 50,000 pictures. The test set contains 300,000 pictures. The image formats in both datasets are PNG, with 32 pixels in height and width, and three color channels (RGB). The images cover 10 categories: airplanes, cars, birds, cats, deer, dogs, frogs, horses, boats and trucks. To make it easier to get started, we provide a small sample of the above data set. "Train_tiny.zip" contains 80 training samples, and "test_tiny.zip" contains 100 test samples. Their uncompressed folder names are "train_tiny" and "test_tiny" respectively.
- Image enhancement
data_transform = transforms.Compose([
transforms.Resize(40),
transforms.RandomHorizontalFlip(),
transforms.RandomCrop(32),
transforms.ToTensor()
])
trainset = torchvision.datasets.ImageFolder(root='/home/kesci/input/CIFAR102891/cifar-10/train'
, transform=data_transform)
trainset[0][0].shape # torch.Size([3, 32, 32])
data = [d[0].data.cpu().numpy() for d in trainset]
np.mean(data) # 0.4676536
np.std(data) # 0.23926772
# 图像增强
transform_train = transforms.Compose([
transforms.RandomCrop(32, padding=4), #先四周填充0，再把图像随机裁剪成32*32
transforms.RandomHorizontalFlip(), #图像一半的概率翻转，一半的概率不翻转
transforms.ToTensor(),
transforms.Normalize((0.4731, 0.4822, 0.4465), (0.2212, 0.1994, 0.2010)), #R,G,B每层的归一化用到的均值和方差
])
transform_test = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.4731, 0.4822, 0.4465), (0.2212, 0.1994, 0.2010)),
])
Import data set
train_dir = '/home/kesci/input/CIFAR102891/cifar-10/train'
test_dir = '/home/kesci/input/CIFAR102891/cifar-10/test'
trainset = torchvision.datasets.ImageFolder(root=train_dir, transform=transform_train)
trainloader = torch.utils.data.DataLoader(trainset, batch_size=256, shuffle=True)
testset = torchvision.datasets.ImageFolder(root=test_dir, transform=transform_test)
testloader = torch.utils.data.DataLoader(testset, batch_size=256, shuffle=False)
classes = ['airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'forg', 'horse', 'ship', 'truck']
- Define the model
ResNet-18 network structure: ResNet full name Residual Network residual network. Kaiming He's "Deep Residual Learning for Image Recognition" won the best paper in CVPR. The deep residual network he proposed in 2015 can be said to have washed out the major image competitions and won multiple competitions with absolute advantages. And under the premise of ensuring the accuracy of the network, the depth of the network reached 152 layers, and later it was further increased to a depth of 1000.
class ResidualBlock(nn.Module): # 我们定义网络时一般是继承的torch.nn.Module创建新的子类
def __init__(self, inchannel, outchannel, stride=1):
super(ResidualBlock, self).__init__()
#torch.nn.Sequential是一个Sequential容器，模块将按照构造函数中传递的顺序添加到模块中。
self.left = nn.Sequential(
nn.Conv2d(inchannel, outchannel, kernel_size=3, stride=stride, padding=1, bias=False),
# 添加第一个卷积层,调用了nn里面的Conv2d（）
nn.BatchNorm2d(outchannel), # 进行数据的归一化处理
nn.ReLU(inplace=True), # 修正线性单元，是一种人工神经网络中常用的激活函数
nn.Conv2d(outchannel, outchannel, kernel_size=3, stride=1, padding=1, bias=False),
nn.BatchNorm2d(outchannel)
)
self.shortcut = nn.Sequential()
if stride != 1 or inchannel != outchannel: #s=1高宽相同 in=out通道数相同
self.shortcut = nn.Sequential(
nn.Conv2d(inchannel, outchannel, kernel_size=1, stride=stride, bias=False),
nn.BatchNorm2d(outchannel)
)
# 便于之后的联合,要判断Y = self.left(X)的形状是否与X相同
def forward(self, x): # 将两个模块的特征进行结合，并使用ReLU激活函数得到最终的特征。(标准残差块的定义)
out = self.left(x)
out += self.shortcut(x)
out = F.relu(out)
return out
class ResNet(nn.Module):
def __init__(self, ResidualBlock, num_classes=10):
super(ResNet, self).__init__()
self.inchannel = 64
self.conv1 = nn.Sequential( # 用3个3x3的卷积核代替7x7的卷积核，减少模型参数
nn.Conv2d(3, 64, kernel_size=3, stride=1, padding=1, bias=False),
nn.BatchNorm2d(64),
nn.ReLU(),
)
self.layer1 = self.make_layer(ResidualBlock, 64, 2, stride=1) #2个残差块
self.layer2 = self.make_layer(ResidualBlock, 128, 2, stride=2)
self.layer3 = self.make_layer(ResidualBlock, 256, 2, stride=2)
self.layer4 = self.make_layer(ResidualBlock, 512, 2, stride=2)
self.fc = nn.Linear(512, num_classes)
def make_layer(self, block, channels, num_blocks, stride):
strides = [stride] + [1] * (num_blocks - 1) #第一个ResidualBlock的步幅由make_layer的函数参数stride指定
# ，后续的num_blocks-1个ResidualBlock步幅是1
layers = []
for stride in strides:
layers.append(block(self.inchannel, channels, stride))
self.inchannel = channels
return nn.Sequential(*layers)
def forward(self, x):
out = self.conv1(x)
out = self.layer1(out)
out = self.layer2(out)
out = self.layer3(out)
out = self.layer4(out)
out = F.avg_pool2d(out, 4)
out = out.view(out.size(0), -1)
out = self.fc(out)
return out
def ResNet18():
return ResNet(ResidualBlock)
Training and testing
I feel that this piece of code is a bit messy, and it is not very comfortable to read. It is purely personal opinion
# 定义是否使用GPU
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
# 超参数设置
EPOCH = 20 #遍历数据集次数
pre_epoch = 0 # 定义已经遍历数据集的次数
LR = 0.1 #学习率
# 模型定义-ResNet
net = ResNet18().to(device)
# 定义损失函数和优化方式
criterion = nn.CrossEntropyLoss() #损失函数为交叉熵，多用于多分类问题
optimizer = optim.SGD(net.parameters(), lr=LR, momentum=0.9, weight_decay=5e-4)
#优化方式为mini-batch momentum-SGD，并采用L2正则化（权重衰减）
# 训练
if __name__ == "__main__":
print("Start Training, Resnet-18!")
num_iters = 0
for epoch in range(pre_epoch, EPOCH):
print('\nEpoch: %d' % (epoch + 1))
net.train()
sum_loss = 0.0
correct = 0.0
total = 0
for i, data in enumerate(trainloader, 0):
#用于将一个可遍历的数据对象(如列表、元组或字符串)组合为一个索引序列，同时列出数据和数据下标，
#下标起始位置为0，返回 enumerate(枚举) 对象。
num_iters += 1
inputs, labels = data
inputs, labels = inputs.to(device), labels.to(device)
optimizer.zero_grad() # 清空梯度
# forward + backward
outputs = net(inputs)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
sum_loss += loss.item() * labels.size(0)
_, predicted = torch.max(outputs, 1) #选出每一列中最大的值作为预测结果
total += labels.size(0)
correct += (predicted == labels).sum().item()
# 每20个batch打印一次loss和准确率
if (i + 1) % 20 == 0:
print('[epoch:%d, iter:%d] Loss: %.03f | Acc: %.3f%% '
% (epoch + 1, num_iters, sum_loss / (i + 1), 100. * correct / total))
print("Training Finished, TotalEPOCH=%d" % EPOCH)