10.初识Pytorch使用池化层并对其进行可视化

游客26024

已于 2022-01-29 00:00:13 修改

阅读量1.1k

点赞数 1

分类专栏：手把手学习Pytorch 文章标签： pytorch 深度学习 python 计算机视觉人工智能

于 2021-12-29 22:45:12 首次发布

本文链接：https://blog.csdn.net/XiaoyYidiaodiao/article/details/122226062

版权

手把手学习Pytorch 专栏收录该内容

23 篇文章 13 订阅

订阅专栏

池化层的编程跟之前的卷积层类似。

最大池化层，上代码：

import torch.nn as nn
import torch
import torchvision
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
from torchvision import transforms

# 创建的单层最大池化层类
# Pytorch中模型的三要素 1.nn.Module
class MyPool(nn.Module):
		# Pytorch中模型的三要素 2.__init__()
    def __init__(self):
        super(MyPool, self).__init__()
        self.maxpool1 = torch.nn.MaxPool2d(kernel_size=3, stride=1, padding=0)
		# Pytorch中模型的三要素 3. forward()
    def forward(self, x):
        x = self.maxpool1(x)
        return x

# 将数据集图像从PIL格式转化为Tensor格式
tran_tensor = transforms.ToTensor()

# 使用CIFAR10数据集
# root数据集的位置,train是否是训练集,transform数据集,download是否下载数据集
dataset = torchvision.datasets.CIFAR10(root="../dataset", train=False, transform=tran_tensor, download=True)

# 加载数据集dataloader
# dataset数据集的位置,batch_size一次性加载的数据集图像,shuffle是否洗牌,num_workers单多线程,drop_last是否舍去最后总数据集个数除以batch_size个数的余数
dataloader = DataLoader(dataset=dataset, batch_size=64, shuffle=True, num_workers=0, drop_last=False)

# 创建Tensorboard.SummaryWriter模板
writer = SummaryWriter("logs")

step = 0
for data in dataloader:
    imgs, targets = data
    print("imgs.shape", imgs.shape)
    writer.add_images("original_imgs", imgs, step)
    # 创建MyPool类
    mypool = MyPool()
    # 将图像喂给单层最大池化层的模型
    outputs = mypool(imgs)
    print("outputs.shape", outputs.shape)
    writer.add_images("outputs", outputs, step)
    step = step + 1
writer.close()

run之后，输入命令行：

tensorboard --logdir=logs

结果：
在这里插入图片描述

平均池化层，上代码：

import torch.nn as nn
import torch
import torchvision
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
from torchvision import transforms


class MyPool(nn.Module):
    def __init__(self):
        super(MyPool, self).__init__()
        self.avgpool1 = torch.nn.AvgPool2d(kernel_size=3, stride=1, padding=0)

    def forward(self, x):
        x = self.avgpool1(x)
        return x


tran_tensor = transforms.ToTensor()
dataset = torchvision.datasets.CIFAR10(root="../dataset", train=False, transform=tran_tensor, download=True)
dataloader = DataLoader(dataset=dataset, batch_size=64, shuffle=True, num_workers=0, drop_last=False)

writer = SummaryWriter("logs")

step = 0
for data in dataloader:
    imgs, targets = data
    print("imgs.shape", imgs.shape)
    writer.add_images("original_imgs", imgs, step)
    mypool = MyPool()
    outputs = mypool(imgs)
    print("outputs.shape", outputs.shape)
    writer.add_images("outputs", outputs, step)
    step = step + 1
writer.close()