【PyTorch】教程：torch.nn.Conv2d

Conv2d

CLASS torch.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)

2D 卷积

$$out(N_i,C_{ou{t}_j})=bias(C_{ou{t}_j})+\sum _{k=0}^{C_{in}-1}Weight(C_{ou{t}_j},k)\ast input(N_i,k)$$

$N$: 批大小
$C$: 通道数
$H$: 输入高
$W$: 输入宽

• weight ([Tensor]) – 模块可学习的权重，形状为 ($\text{out\_chanels}$, $\frac{\text{in\_channels}}{\text{groups}}$, $\text{kernel\_size[0]}$, $\text{kernel\_size[1]}$). 从 $\mathcal{U}(-\sqrt{k},\sqrt{k})$ 中采样，其中$k=\frac{groups}{C_{\text{in}}\ast {\prod }_{i=0}^1\text{kernel\_size}[i]}$
• bias ([Tensor] ) – 模块可学习的权重，形状为 (out_channels). 如果 bias 为 True, 从 $\mathcal{U}(-\sqrt{k},\sqrt{k})$ 采样，其中$k=\frac{groups}{C_{\text{in}}\ast {\prod }_{i=0}^1\text{kernel\_size}[i]}$

• 参数
• in_channels ([int]) - 输入图像的通道数
• out_channels ([int]) - 输出通道数
• kernel_size ([int] or *[tuple]) - 卷积核尺寸
• stride ([int] or [tuple], optional ) - 步长，默认为1
• padding ([int] [tuple] or [str] optional ) - 扩边，默认为0
• padding_mode ([str] optional ) - zeros, reflect, replicate, 默认为 zeros.
• dilation ([int] [tuple] optional ) - 膨胀核，默认为1
• groups ([int] optional ) - 分组，默认为1
• bias ([bool] optional ) - 如果设置为 true, 添加学习的 bias

import torch
import torch.nn as nn 

# With square kernels and equal stride
m1 = nn.Conv2d(16, 33, 3, stride=2)
# non-square kernels and unequal stride and with padding
m2 = nn.Conv2d(16, 33, (3, 5), stride=(2, 1), padding=(4, 2))
# non-square kernels and unequal stride and with padding and dilation
m3 = nn.Conv2d(16, 33, (3, 5), stride=(2, 1), padding=(4, 2), dilation=(3, 1))

input = torch.randn(20, 16, 50, 100)

output1 = m1(input)  # torch.Size([20, 33, 24, 49])
output2 = m2(input)  # torch.Size([20, 33, 28, 100])
output3 = m3(input)  # torch.Size([20, 33, 26, 100])

【参考】

Conv2d — PyTorch 1.13 documentation