卷积初步认识

import torch
from torch import nn
"""
[N,C,H,W]
N：批量大小
C：通道数
H：高
W：宽
nn.Conv2d(
    in_channels: int,  输入图像的通道数
    out_channels: int, 输出图像的通道数
    kernel_size: Union[int, Tuple[int, int]],卷积核
    stride: Union[int, Tuple[int, int]] = 1, 步长
    padding: Union[str, int, Tuple[int, int]] = 0, 填充，补零
    dilation: Union[int, Tuple[int, int]] = 1,膨胀系数
    groups: int = 1, 分组卷积
    bias: bool = True, 偏置
    padding_mode: str = 'zeros',填充模式
    device=None,设备 cpu 或者gpu
    dtype=None,数据的类型，默认 torch.float32
) -> None

"""
conv1 = nn.Conv2d(in_channels=3,out_channels=16,kernel_size=3,stride=1,padding=1)
# 一个图片，3层网络，64x64大小
X = torch.randn(1,3,64,64)

conv1(X).shape
# 输出：1个图像，16层网络，大小64x64
#torch.Size([1, 16, 64, 64])