7、mobilenetv1

import torch
import torch.nn as nn


# 深度可分离卷积
class Depth_Separable_Conv(nn.Module):

    def __init__(self, in_channels, out_channels, stride=1):
        super().__init__()
        self.in_channels = in_channels
        self.out_channels = out_channels
        self.stride = stride

        self.dw_conv = self.depth_wise_conv()  # 逐通道卷积：groups=in_channels
        self.pw_conv = self.point_wise_conv()  # 逐点卷积，即普通1x1卷积

    # 逐通道卷积：groups=in_channels
    def depth_wise_conv(self):
        return nn.Sequential(
            nn.Conv2d(in_channels=self.in_channels, out_channels=self.in_channels, kernel_size=3, stride=self.stride,
                      padding=1, groups=self.in_channels, bias=False),
            nn.BatchNorm2d(self.in_channels),
            nn.ReLU()
        )

    # 逐点卷积，即普通1x1卷积
    def point_wise_conv(self):
        return nn.Sequential(
            nn.Conv2d(in_channels=self.in_channels, out_channels=self.out_channels, kernel_size=1, stride=1, padding=0,
                      groups=1, bias=False),
            nn.BatchNorm2d(self.out_channels),
            nn.ReLU()
        )

    def forward(self, x):
        x = self.dw_conv(x)
        x = self.pw_conv(x)
        return x


class MobileNetV1(nn.Module):

    def __init__(self, num_classes=1000, img_channels=3):
        super().__init__()
        # 输出类别数
        self.num_classes = num_classes
        # 输入图片的通道数
        self.img_channels = img_channels
        # 网络结构
        self.m = nn.Sequential(
            nn.Conv2d(in_channels=self.img_channels, out_channels=32, kernel_size=3, stride=2, padding=1, groups=1,
                      bias=False),
            nn.BatchNorm2d(32),
            nn.ReLU(),

            # Depth_Separable_Conv的参数顺序：in_channels,out_channels,stride
            Depth_Separable_Conv(32, 64, 1),
            Depth_Separable_Conv(64, 128, 2),
            Depth_Separable_Conv(128, 128, 1),
            Depth_Separable_Conv(128, 256, 2),
            Depth_Separable_Conv(256, 256, 1),
            Depth_Separable_Conv(256, 512, 2),

            # 重复5次
            Depth_Separable_Conv(512, 512, 1),
            Depth_Separable_Conv(512, 512, 1),
            Depth_Separable_Conv(512, 512, 1),
            Depth_Separable_Conv(512, 512, 1),
            Depth_Separable_Conv(512, 512, 1),

            Depth_Separable_Conv(512, 1024, 2),
            Depth_Separable_Conv(1024, 1024, 2),

            nn.AdaptiveAvgPool2d((1, 1)),
            nn.Flatten(start_dim=1),
            nn.Linear(in_features=1024, out_features=self.num_classes)
        )

    def forward(self, x):
        x = self.m(x)
        return x
if __name__=='__main__':
    model = MobileNetV1()
    print(model)