PyTorch——深度可分離卷積（一）

1、深度可分離卷積

Depthwise Separable Convolution

（一）結構

實質上是將標准卷積分成了兩步：depthwise卷積和pointwise卷積。

標准卷積：

depthwise卷積：

pointwise卷積：

2、代碼實現

[32, 3, 224, 224] ——> [32, 64, 112, 112]

import torch
import torch.nn as nn
import torch.nn.functional as F

class Block(nn.Module):
    "Depthwise conv + Pointwise conv"
    def __init__(self, in_channels, out_channels, stride=1):
        super(Block, self).__init__()
        self.conv1 = nn.Conv2d(in_channels, in_channels, kernel_size=3, stride=2, padding=1, groups=in_channels, bias=False)
        self.bn1 = nn.BatchNorm2d(in_channels)
        self.conv2 = nn.Conv2d(in_channels, out_channels, kernel_size=1, stride=1, padding=0, bias=False)
        self.bn2 = nn.BatchNorm2d(out_channels)

    def forward(self, x):
        x = self.conv1(x)
        x = self.bn1(x)
        x = F.relu(x)
        x = self.conv2(x)
        x = self.bn2(x)
        x = F.relu(x)
        return x

input = torch.randn(32, 3, 224, 224)
block = Block(3, 64)
out = block(input)
print(out.size())