Conv2d实例

import torch
import torchvision
from torch import nn
from torch.nn import Conv2d
from torch.utils.data import DataLoader
# 必须转换成Tensor的数据类型，才能送到网络中
from torch.utils.tensorboard import SummaryWriter

transform = torchvision.transforms.ToTensor()
train_data=torchvision.datasets.CIFAR10('dataset',train=True,transform=None,download=True)
test_data = torchvision.datasets.CIFAR10('dataset',train=False,transform=transform,download=True)
# 分成几个batch
test_loader = DataLoader(test_data,batch_size=64)

# 搭建神经网络框架,重写Moudle类
class Moudle(nn.Module):
    def __init__(self):
        super().__init__()
        # self.的方式就可以在其他的方法中也使用该变量
        ####
        # 原来是32*32的图像，如果没有用padding，卷积后会变成30*30的图像
        self.conv1 = Conv2d(in_channels=3,out_channels=6,kernel_size=(3,3),stride=1,padding=0)
    def forward(self, input):
        output = self.conv1(input)
        return output

writer = SummaryWriter('logs')
# 初始化对象
moduel = Moudle()
# 对每一个batch中的图像都进行卷积操作
step = 0
for data in test_loader:
    images ,target =data
    # torch.size([64, 6, 32, 32])
    output = moduel(images)
    writer.add_images('test',images,global_step=step,dataformats='NCHW')
    # torch.size([64,6,30,30])无法显示，reshape为(xxx,3,30,30)，不知道就填-1
    #####
    #注意，这边reshape完后，必须把这个矩阵重新赋值给一个新的矩阵，torch.resahpe并不会改变原矩阵的大小和形状
    output =torch.reshape(output,(-1,3,30,30))
    print(output.shape)
    writer.add_images('conv', output, global_step=step, dataformats='NCHW')
    step =step+1

transform举例

transform = transforms.Compose([
        transforms.CenterCrop(224),
        transforms.Resize((224, 224)),
        transforms.ToTensor(),
        transforms.Normalize(mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5))  # 归一化
    ])