Maxpool2d explanation -- Application in arrays and images

1、 Environmental requirements

1、 Need to install Pytorch rely on
2、 Official documents conv2d
3、 Picture needs CIFAR10 Data sets

2、 Principle explanation

Cover the original data with convolution kernel , Select the maximum value covered by convolution kernel in the original data

Select the maximum value of convolution kernel coverage ,ceil_mode Controls whether the convolution kernel is retained after exceeding the original data

3、 Function requirements

function ：

Parameter requirements

kernel_size Set convolution kernel size attribute
stride and conv2d Medium stride equally , Is an attribute that controls the movement stride , Note here ,conv2d The default value is 1, however MaxPool2d The default value is the convolution kernel size
padding Set the properties filled around the original data
dilation: Indicates the addition of... Between the original data 0 Properties of
ceil_mode Controls when the convolution kernel exceeds the original image , Whether the max Make a reservation

4、 Example

4.1、 Array

Code ：

import torch
import torchvision

from torch.nn import Module,MaxPool2d
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter

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)

print(" front ",input.shape)      # torch.Size([5, 5]), The input conditions are not met , Format conversion required 

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

print(" after ",input.shape)    #  after  torch.Size([1, 1, 5, 5])  One bach_size,

class ConNet(Module):
    def __init__(self):
        super(ConNet, self).__init__()
        #  Use of pool layer , Set convolution kernel to 3*3, The excess part retains the data 
        self.maxpool = MaxPool2d(kernel_size=3,ceil_mode=True)

    def forward(self,input):
        output = self.maxpool(input)
        return output

#  Instantiate objects 
Work = ConNet()

#  Neural network call 
output = Work(input)
print(output)

result ：

4.2、 Images

Code ：

import torch
import torchvision

from torch.nn import Module,MaxPool2d
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter

class ConNet(Module):
    def __init__(self):
        super(ConNet, self).__init__()
        #  Use of pool layer , Set convolution kernel to 3*3, The excess part retains the data 
        self.maxpool = MaxPool2d(kernel_size=3,ceil_mode=True)

    def forward(self,input):
        output = self.maxpool(input)
        return output

#  Instantiate objects 
Work = ConNet()

# CIFAR10 Data usage 
dataset = torchvision.datasets.CIFAR10("./datasetvision",train=False,download=False,transform=torchvision.transforms.ToTensor())

#  Data loading 
dataloader = DataLoader(dataset,batch_size=64)

writer = SummaryWriter("logs_MaxPool")

step = 0
for data in dataloader:
    imgs,target = data
    writer.add_images("input",imgs,step)
    output = Work(imgs)
    writer.add_images("output",output,step)
    step = step + 1

writer.close()

result ：

4.3、Conv2d+MaxPool2d Images

Code ：

import torch
import torchvision

from torch.nn import Module,MaxPool2d,Conv2d
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter

class ConNet(Module):
    def __init__(self):
        super(ConNet, self).__init__()
        #  Use of pool layer , Set convolution kernel to 3*3, The excess part retains the data 
        self.maxpool = MaxPool2d(kernel_size=3,ceil_mode=True)
        self.conv2d = Conv2d(in_channels=3,out_channels=3,kernel_size=3,stride=1,padding=0)

    def forward(self,input):
        output = self.conv2d(input)
        output = self.maxpool(output)
        return output

#  Instantiate objects 
Work = ConNet()

# CIFAR10 Data usage 
dataset = torchvision.datasets.CIFAR10("./datasetvision",train=False,download=False,transform=torchvision.transforms.ToTensor())

#  Data loading 
dataloader = DataLoader(dataset,batch_size=64)

writer = SummaryWriter("logs_MaxPoolAndConv2d")

step = 0
for data in dataloader:
    imgs,target = data
    writer.add_images("input",imgs,step)
    output = Work(imgs)
    writer.add_images("output",output,step)
    step = step + 1

writer.close()

result ：