本文主要是介绍【超级干货】2天速成PyTorch深度学习入门教程，缓解研究生焦虑，希望对大家解决编程问题提供一定的参考价值，需要的开发者们随着小编来一起学习吧！
3、cnn基础
卷积神经网络
输入层 —输入图片矩阵
- 输入层一般是 RGB 图像或单通道的灰度图像，图片像素值在[0,255]，可以用矩阵表示图片
卷积层 —特征提取
- 人通过特征进行图像识别，根据左图直的笔画判断X，右图曲的笔画判断圆
- 卷积操作
激活层 —加强特征
池化层 —压缩数据
全连接层 —进行分类
输出层 —输出分类概率
4、基于LeNet实现cifar10数据集分类
1、数据集
- DataSet
import torchvision
from torchvision import transforms
def get_train_dataset(data_root):transform = transforms.Compose([transforms.ToTensor(),#0~1transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])#-1,1#(0-0.5)/0.5=-1 (1-0.5)/0.5=1train_dataset = torchvision.datasets.CIFAR10(root=data_root, train=True,download=True, transform=None)return train_dataset
train_dataset = get_train_dataset("dataset")
image,label = train_dataset[1]
print(train_dataset.classes[label])
print(type(image))
transforms 设定处理图片的规则
# Composes several transforms together (authority)
# 把一系列图片操作【组合】起来
transform = transforms.Compose
# Convert a ``PIL Image`` or ``numpy.ndarray`` to tensor
# 将PIL Image格式或者numpy.ndarray格式的数据格式化为可被pytorch快速处理的tensor类型
transforms.ToTensor()
# Normalize a tensor image with mean and standard deviation
# output[channel] = (input[channel] - mean[channel]) / std[channel]
# 使用均值和方差对数据归一化，保证程序运行时收敛加快，训练次数少点
# 因为tensor是0-1，经过normalize可以变成(-1,1) 所以可以使用0.5，当然并不一定都是0.5
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
# root ：Root directory of dataset where directory(数据集的根目录)
# train If True, creates dataset from training set
# download If true, downloads the dataset from the internet
# transform A transform that takes in an PIL image and returns a transformed version
torchvision.datasets.CIFAR10(root=data_root, train=True,download=True, transform=transform)
验证检查
train_dataset = get_train_dataset("dataset")
image,label = train_dataset[1]
print(train_dataset.classes[label])
image.show()- DataLoader
train_loader = torch.utils.data.DataLoader(dataset=train_dataset, batch_size=32,shuffle=True, num_workers=2)说明
dataset:数据集
batch_size:how many samples per batch to load
shuffle:set to ``True`` to have the data reshuffled
num_workers:
how many subprocesses to use for data 加载数据集采用单进程还是多进程
0 means that the data will be loaded in the main process.数据在主进程加载（windows建议0）测试
val_dataset = get_val_dataset("dataset")
val_loader = get_val_loader(val_dataset)
for data in val_loader:images,labels = dataprint(images.shape)print(labels)
2、模型
- 搭建网络
import torch
from torch import nnclass Net(nn.Module):def __init__(self):super().__init__()self.conv1 = nn.Conv2d(in_channels=3, out_channels=6,kernel_size=5,stride=1,padding=0)self.relu1 = nn.ReLU()self.pool1 = nn.MaxPool2d(kernel_size=2, stride=2)self.conv2 = nn.Conv2d(in_channels=6, out_channels=16,kernel_size=5,stride=1,padding=0)self.relu2 = nn.ReLU()self.pool2 = nn.MaxPool2d(kernel_size=2, stride=2)self.flatten = nn.Flatten()self.fc1 = nn.Linear(in_features=16 * 5 * 5, out_features=120)self.relu3 = nn.ReLU()self.fc2 = nn.Linear(in_features=120, out_features=84)self.relu4 = nn.ReLU()self.fc3 = nn.Linear(in_features=84, out_features=10)def forward(self, x):x = self.conv1(x)x = self.relu1(x)x = self.pool1(x)x = self.conv2(x)x = self.relu2(x)x = self.pool2(x)x = self.flatten(x)print(x.shape)x = self.fc1(x)x = self.relu3(x)x = self.fc2这篇关于【超级干货】2天速成PyTorch深度学习入门教程，缓解研究生焦虑的文章就介绍到这儿，希望我们推荐的文章对编程师们有所帮助！
