10.nn_maxpooling

https://pytorch.org/docs/stable/generated/torch.nn.MaxPool2d.html#torch.nn.MaxPool2d

torch.nn.MaxPool2d(kernel_size, stride=None, padding=0, dilation=1, return_indices=False, ceil_mode=False)

ceil_mode:如上图这种情况，3×3的池化层只能覆盖2×2的位置（未完全覆盖时），结果是否保留是由ceil_mode来确定，True则保留，False则不保留

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)
# Maxpool2d's Input: :math:`(N, C, H_{in}, W_{in})
input = torch.reshape(input, (-1, 1, 5, 5))
print(input.shape)
class DEMO(nn.Module):
    def __init__(self):
        super(DEMO, self).__init__()
        self.maxpool1 = MaxPool2d(kernel_size=3,ceil_mode=False)
    def forward(self,x):
        x = self.maxpool1(x)
        return x
demo = DEMO()
output = demo(input)
print(output)

ceil_mode设置成True时：

ceil_mode设置成False时：

以CIFAR10数据集为例：

池化不同于卷积层，卷积层要设置in_channels 和 out_channels，池化时 输入为三通道的RGB图片输出仍然是三通道的图片。池化的作用可以理解为提取出特征 ，减小运算量。

import torch
from torch import nn
from torch.nn import MaxPool2d
import torchvision
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
dataset = torchvision.datasets.CIFAR10('./dataset',train=False,transform=torchvision.transforms.ToTensor(),
                                       download=True)
dataloader = DataLoader(dataset,batch_size=64)
class DEMO(nn.Module):
    def __init__(self):
        super(DEMO, self).__init__()
        self.maxpool1 = MaxPool2d(kernel_size=3,ceil_mode=False)
    def forward(self,x):
        x = self.maxpool1(x)
        return x
demo = DEMO()
writer = SummaryWriter('./logs')
step = 0
for data in dataloader:
    imgs,targets = data
    output = demo(imgs)
    print(imgs.shape)
    print(output.shape)
    writer.add_images('input_maxpool',imgs,step)
    writer.add_images('output_maxpool',output,step)
writer.close()