File size: 5,787 Bytes
bdbd148
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
'''
MobileNetV2 in PyTorch.

论文: "Inverted Residuals and Linear Bottlenecks: Mobile Networks for Classification, Detection and Segmentation"
参考: https://arxiv.org/abs/1801.04381

主要特点:
1. 引入倒残差结构(Inverted Residual),先升维后降维
2. 使用线性瓶颈(Linear Bottlenecks),去除最后一个ReLU保留特征
3. 使用ReLU6作为激活函数,提高在低精度计算下的鲁棒性
4. 残差连接时使用加法而不是拼接,减少内存占用
'''

import torch
import torch.nn as nn


class Block(nn.Module):
    '''倒残差块 (Inverted Residual Block)
    
    结构: expand(1x1) -> depthwise(3x3) -> project(1x1)
    特点:
    1. 使用1x1卷积先升维再降维(与ResNet相反)
    2. 使用深度可分离卷积减少参数量
    3. 使用shortcut连接(当stride=1且输入输出通道数相同时)
    
    Args:
        in_channels: 输入通道数
        out_channels: 输出通道数
        expansion: 扩展因子,控制中间层的通道数
        stride: 步长,控制特征图大小
    '''
    def __init__(self, in_channels, out_channels, expansion, stride):
        super(Block, self).__init__()
        self.stride = stride
        channels = expansion * in_channels  # 扩展通道数
        
        # 1x1卷积升维
        self.conv1 = nn.Conv2d(
            in_channels, channels, 
            kernel_size=1, stride=1, padding=0, bias=False
        )
        self.bn1 = nn.BatchNorm2d(channels)
        
        # 3x3深度可分离卷积
        self.conv2 = nn.Conv2d(
            channels, channels,
            kernel_size=3, stride=stride, padding=1, 
            groups=channels, bias=False  # groups=channels即为深度可分离卷积
        )
        self.bn2 = nn.BatchNorm2d(channels)
        
        # 1x1卷积降维(线性瓶颈,不使用激活函数)
        self.conv3 = nn.Conv2d(
            channels, out_channels,
            kernel_size=1, stride=1, padding=0, bias=False
        )
        self.bn3 = nn.BatchNorm2d(out_channels)
        
        # shortcut连接
        self.shortcut = nn.Sequential()
        if stride == 1 and in_channels != out_channels:
            self.shortcut = nn.Sequential(
                nn.Conv2d(
                    in_channels, out_channels,
                    kernel_size=1, stride=1, padding=0, bias=False
                ),
                nn.BatchNorm2d(out_channels)
            )
            
        self.relu6 = nn.ReLU6(inplace=True)
    
    def forward(self, x):
        # 主分支
        out = self.relu6(self.bn1(self.conv1(x)))  # 升维
        out = self.relu6(self.bn2(self.conv2(out)))  # 深度卷积
        out = self.bn3(self.conv3(out))  # 降维(线性瓶颈)
        
        # shortcut连接(仅在stride=1时)
        out = out + self.shortcut(x) if self.stride == 1 else out
        return out


class MobileNetV2(nn.Module):
    '''MobileNetV2网络
    
    Args:
        num_classes: 分类数量
    
    网络配置:
        cfg = [(expansion, out_channels, num_blocks, stride), ...]
        - expansion: 扩展因子
        - out_channels: 输出通道数
        - num_blocks: 块的数量
        - stride: 第一个块的步长
    '''
    # 网络结构配置
    cfg = [
        # (expansion, out_channels, num_blocks, stride)
        (1,  16, 1, 1),  # conv1
        (6,  24, 2, 1),  # conv2,注意:原论文stride=2,这里改为1以适应CIFAR10
        (6,  32, 3, 2),  # conv3
        (6,  64, 4, 2),  # conv4
        (6,  96, 3, 1),  # conv5
        (6, 160, 3, 2),  # conv6
        (6, 320, 1, 1),  # conv7
    ]
    
    def __init__(self, num_classes=10):
        super(MobileNetV2, self).__init__()
        
        # 第一层卷积(注意:原论文stride=2,这里改为1以适应CIFAR10)
        self.conv1 = nn.Conv2d(3, 32, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn1 = nn.BatchNorm2d(32)
        
        # 主干网络
        self.layers = self._make_layers(in_channels=32)
        
        # 最后的1x1卷积
        self.conv2 = nn.Conv2d(320, 1280, kernel_size=1, stride=1, padding=0, bias=False)
        self.bn2 = nn.BatchNorm2d(1280)
        
        # 分类器
        self.avgpool = nn.AdaptiveAvgPool2d(1)  # 全局平均池化
        self.linear = nn.Linear(1280, num_classes)
        self.relu6 = nn.ReLU6(inplace=True)
        
    def _make_layers(self, in_channels):
        '''构建网络层
        
        Args:
            in_channels: 输入通道数
        '''
        layers = []
        for expansion, out_channels, num_blocks, stride in self.cfg:
            # 对于每个配置,第一个block使用指定的stride,后续blocks使用stride=1
            strides = [stride] + [1]*(num_blocks-1)
            for stride in strides:
                layers.append(
                    Block(in_channels, out_channels, expansion, stride)
                )
                in_channels = out_channels
        return nn.Sequential(*layers)
    
    def forward(self, x):
        # 第一层卷积
        out = self.relu6(self.bn1(self.conv1(x)))
        
        # 主干网络
        out = self.layers(out)
        
        # 最后的1x1卷积
        out = self.relu6(self.bn2(self.conv2(out)))
        
        # 分类器
        out = self.avgpool(out)
        out = out.view(out.size(0), -1)
        out = self.linear(out)
        return out


def test():
    """测试函数"""
    net = MobileNetV2()
    x = torch.randn(2, 3, 32, 32)
    y = net(x)
    print(y.size())
    
    # 打印模型结构
    from torchinfo import summary
    device = 'cuda' if torch.cuda.is_available() else 'cpu'
    net = net.to(device)
    summary(net, (2, 3, 32, 32))

if __name__ == '__main__':
    test()