神经网络：池化层

神经网络池化操作

下采样：减少特征数量

先看池化操作：

其中ceil_mode设置成True和False池化区别：

在这个例子当中：ceil_mode=True表示边缘不满3x3的部分也会被池化，False表示边缘不满3x3的部分不会采样

code：

python 复制代码

import torch
from torch import nn
from torch.nn import MaxPool2d

input = torch.tensor([[1,2,0,3,1],
                      [0,1,2,3,1],
                      [1,2,1,0,0],
                      [5,2,3,1,1],
                      [2,1,0,1,1]
],dtype = torch.float32)
#这里dtype为float是因为maxpool2d只能处理float类型的数据

input = torch.reshape(input,(-1,1,5,5))
print(input.shape)

class Net(nn.Module):
    def __init__(self):
        super(Net,self).__init__()
        #ceil_mode=True表示边缘不满3x3的部分也会被池化
        #kernel_size=3 默认是卷积核的大小
        self.maxpool1 = MaxPool2d(kernel_size=3,ceil_mode=True)
        self.maxpool2 = MaxPool2d(kernel_size=3,ceil_mode=False)

    def forward(self,input):
        #output = self.maxpool1(input)
        output = self.maxpool2(input)
        return output

net = Net()
output = net(input)
print(output)

ceil_mode=True:

ceil_mode=False:

神经网络池化层

这里需要先看这篇博客：

https://blog.csdn.net/whdehcy/article/details/149486555?fromshare=blogdetail\&sharetype=blogdetail\&sharerId=149486555\&sharerefer=PC\&sharesource=whdehcy\&sharefrom=from_link

是讲卷积层的

现在将上一步的卷积得到的特征图作为池化的输入

python 复制代码

    pool_output = poolnet(conv_output)
    writer.add_images('pool_output',pool_output,cnt)

只需要添加一下池化的操作

python 复制代码

class poolNet(nn.Module):
    def __init__(self):
        super(poolNet,self).__init__()
        #ceil_mode=True表示边缘不满3x3的部分也会被池化
        #kernel_size=3 默认是卷积核的大小
        self.maxpool1 = MaxPool2d(kernel_size=3,ceil_mode=True)
        self.maxpool2 = MaxPool2d(kernel_size=3,ceil_mode=False)

    def forward(self,input):
        output = self.maxpool1(input)
        #output = self.maxpool2(input)
        return output

poolnet = poolNet()

完整版代码：

python 复制代码

import torch
import torchvision
from torch import nn
from torch.nn import Conv2d, MaxPool2d
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
from torchvision.datasets import ImageFolder
from torchvision import transforms

#数据预处理
transform = transforms.Compose([
    transforms.Resize((224,224)),
    transforms.ToTensor(),
    transforms.Normalize(
        mean = [0.5,0.5,0.5],
        std = [0.5,0.5,0.5]
    )
])

#加载数据集
folder_path = '../images'
dataset = ImageFolder(folder_path,transform=transform)
dataloader = DataLoader(dataset,batch_size=1)

#卷积
class convNet(nn.Module):
    def __init__(self):
        #调用父类nn.Module的构造函数
        super(convNet,self).__init__()
        self.conv1 = Conv2d(in_channels=3,out_channels=6,kernel_size=3,stride=1,padding=0)

    def forward(self,x):
        x = self.conv1(x)
        return x

convnet = convNet()

#池化
class poolNet(nn.Module):
    def __init__(self):
        super(poolNet,self).__init__()
        #ceil_mode=True表示边缘不满3x3的部分也会被池化
        #kernel_size=3 默认是卷积核的大小
        self.maxpool1 = MaxPool2d(kernel_size=3,ceil_mode=True)
        self.maxpool2 = MaxPool2d(kernel_size=3,ceil_mode=False)

    def forward(self,input):
        output = self.maxpool1(input)
        #output = self.maxpool2(input)
        return output

poolnet = poolNet()

writer = SummaryWriter('../logs')

cnt = 0
for data in dataloader:
    img,label = data
    print(img.shape)
    conv_output = convnet(img)
    print(conv_output.shape)
    writer.add_images('input',img,cnt)
    conv_output = torch.reshape(conv_output,(-1,3,222,222))
    writer.add_images('conv_output',conv_output,cnt)
    pool_output = poolnet(conv_output)
    writer.add_images('pool_output',pool_output,cnt)
    cnt = cnt + 1

writer.close()

卷积：

池化：