1. 池化层原理¶
① 最大池化层有时也被称为下采样。
② dilation为空洞卷积，如下图所示。
③ Ceil_model为当超出区域时，只取最左上角的值。
④ 池化使得数据由5 * 5 变为3 * 3,甚至1 * 1的，这样导致计算的参数会大大减小。例如1080P的电影经过池化的转为720P的电影、或360P的电影后，同样的网速下，视频更为不卡。
2. 池化层处理数据¶
In [1]:
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)
input = torch.reshape(input,(-1,1,5,5))
print(input.shape)
class Tudui(nn.Module):
def __init__(self):
super(Tudui, self).__init__()
self.maxpool = MaxPool2d(kernel_size=3, ceil_mode=True)
def forward(self, input):
output = self.maxpool(input)
return output
tudui = Tudui()
output = tudui(input)
print(output)
torch.Size([1, 1, 5, 5])
tensor([[[[2., 3.],
[5., 1.]]]])
3. 池化层处理图片¶
In [2]:
import torch
import torchvision
from torch import nn
from torch.nn import MaxPool2d
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 Tudui(nn.Module):
def __init__(self):
super(Tudui, self).__init__()
self.maxpool = MaxPool2d(kernel_size=3, ceil_mode=True)
def forward(self, input):
output = self.maxpool(input)
return output
tudui = Tudui()
writer = SummaryWriter("logs")
step = 0
for data in dataloader:
imgs, targets = data
writer.add_images("input", imgs, step)
output = tudui(imgs)
writer.add_images("output", output, step)
step = step + 1
Files already downloaded and verified
① 在 Anaconda 终端里面，激活py3.6.3环境，再输入 tensorboard --logdir=C:\Users\wangy\Desktop\03CV\logs 命令，将网址赋值浏览器的网址栏，回车，即可查看tensorboard显示日志情况。
