【Pytorch】神经网络-卷积层 - 学习笔记

视频地址
打开torch官方文档的卷积层页面

最常用的是nn.conv2d，点击

CLASStorch.nn.Conv2d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros', device=None, dtype=None)

一些不太清楚的

• groups：分组卷积，几乎用不到
• bias：通常设置为True

写代码初始化一下model

import torch
import torchvision
from torch import nn
from torch.nn import Conv2d
from torch.utils.data import DataLoader

dataset = torchvision.datasets.CIFAR10("./dataset", train=False, transform=torchvision.transforms.ToTensor())

dataloader = DataLoader(dataset, batch_size=64)

class Model(nn.Module):
    def __init__(self):
        super(Model, self).__init__()
        self.conv1 = Conv2d(in_channels=3, out_channels=6, kernel_size=3, stride=1, padding=0)  # 把self后面的变量在其它函数里也能够用到

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

model = Model()
print(model)

输出结果为

D:\Anaconda3\envs\pytorch\python.exe D:/研究生/代码尝试/nn_conv2d.py
Model(
  (conv1): Conv2d(3, 6, kernel_size=(3, 3), stride=(1, 1))
)

进程已结束，退出代码为 0

再把dataloder装进去，即图片输入

for data in dataloader:
    imgs, targets = data
    output = model(imgs)
    print(imgs.shape)
    print(output.shape)

输出结果为（前两行）

torch.Size([64, 3, 32, 32])
torch.Size([64, 6, 30, 30])

可以看出，这里的batch_size是64，输入channel有3层，卷积后有6个channel，卷积核不需要初始化，是自动初始化的

可视化，用Tensorboard显示一下，顺便把代码贴全

import torch
import torchvision
from torch import nn
from torch.nn import Conv2d
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter

dataset = torchvision.datasets.CIFAR10("./dataset", train=False, transform=torchvision.transforms.ToTensor())

dataloader = DataLoader(dataset, batch_size=64)

class Model(nn.Module):
    def __init__(self):
        super(Model, self).__init__()
        self.conv1 = Conv2d(in_channels=3, out_channels=6, kernel_size=3, stride=1, padding=0)  # 把self后面的变量在其它函数里也能够用到

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

model = Model()

writer = SummaryWriter("./logs")
step = 0
for data in dataloader:
    imgs, targets = data
    output = model(imgs)
    print(imgs.shape)
    print(output.shape)
    # 输入大小 torch.Size([64, 3, 32, 32])
    writer.add_images("input", imgs, step)
    # 输出大小 torch.Size([64, 6, 30, 30]) 但是六个channel没法显示，所以要转换一下
    # ([64, 6, 30, 30]) ->([xxx, 3, 30, 30])
    output = torch.reshape(output, (-1, 3, 30, 30))
    writer.add_images("output", output, step)
    step = step + 1

还是老样子

tensorboard --logdir=logs

点开后。。。麦艾斯。。。


因为通道从6个压缩成了3个，所以这里有128张图