1
import torch.nn as nn
2
import math
3

4

5
def conv3x3(in_planes, out_planes, stride=1):
6
    """3x3 convolution with padding"""
7
    return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,
8
                     padding=1, bias=False)
9

10

11
class BasicBlock(nn.Module):
12
    expansion = 1
13

14
    def __init__(self, inplanes, planes, stride=1, downsample=None):
15
        super(BasicBlock, self).__init__()
16
        self.conv1 = conv3x3(inplanes, planes, stride)
17
        self.bn1 = nn.BatchNorm2d(planes)
18
        self.relu = nn.ReLU(inplace=True)
19
        self.conv2 = conv3x3(planes, planes)
20
        self.bn2 = nn.BatchNorm2d(planes)
21
        self.downsample = downsample
22
        self.stride = stride
23

24
    def forward(self, x):
25
        residual = x
26

27
        out = self.conv1(x)
28
        out = self.bn1(out)
29
        out = self.relu(out)
30

31
        out = self.conv2(out)
32
        out = self.bn2(out)
33

34
        if self.downsample is not None:
35
            residual = self.downsample(x)
36

37
        out += residual
38
        out = self.relu(out)
39

40
        return out
41

42

43
class Bottleneck(nn.Module):
44
    expansion = 4
45

46
    def __init__(self, inplanes, planes, stride=1, downsample=None):
47
        super(Bottleneck, self).__init__()
48
        self.conv1 = nn.Conv2d(inplanes, planes, kernel_size=1, stride=stride, bias=False)
49
        self.bn1 = nn.BatchNorm2d(planes)
50
        self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=1, padding=1, bias=False)
51
        self.bn2 = nn.BatchNorm2d(planes)
52
        self.conv3 = nn.Conv2d(planes, planes * 4, kernel_size=1, bias=False)
53
        self.bn3 = nn.BatchNorm2d(planes * 4)
54
        self.relu = nn.ReLU(inplace=True)
55
        self.downsample = downsample
56
        self.stride = stride
57

58
    def forward(self, x):
59
        residual = x
60

61
        out = self.conv1(x)
62
        out = self.bn1(out)
63
        out = self.relu(out)
64

65
        out = self.conv2(out)
66
        out = self.bn2(out)
67
        out = self.relu(out)
68

69
        out = self.conv3(out)
70
        out = self.bn3(out)
71

72
        if self.downsample is not None:
73
            residual = self.downsample(x)
74

75
        out += residual
76
        out = self.relu(out)
77

78
        return out
79

80

81
class ResNet(nn.Module):
82

83
    def __init__(self, block, layers, num_classes=1000, include_top=True):
84
        self.inplanes = 64
85
        super(ResNet, self).__init__()
86
        self.include_top = include_top
87
        
88
        self.conv0 = nn.Conv2d(3, 64, kernel_size=7, stride=1, padding=3, bias=False)
89
        self.bn0 = nn.BatchNorm2d(64)
90
        self.relu0 = nn.ReLU(inplace=True)
91
        
92
        self.conv1 = nn.Conv2d(64, 64, kernel_size=7, stride=2, padding=3, bias=False)
93
        self.bn1 = nn.BatchNorm2d(64)
94
        self.relu = nn.ReLU(inplace=True)
95

96
        self.layer1 = self._make_layer(block, 32, layers[0], stride=2)
97
        self.layer2 = self._make_layer(block, 64, layers[1], stride=2)
98
        self.layer3 = self._make_layer(block, 128, layers[2], stride=2)
99
        self.layer4 = self._make_layer(block, 256, layers[3], stride=2)
100
        self.layer5 = self._make_layer(block, 512, layers[4], stride=2)
101
        self.layer6 = self._make_layer(block, 256, layers[5], stride=2)
102

103
        for m in self.modules():
104
            if isinstance(m, nn.Conv2d):
105
                n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
106
                m.weight.data.normal_(0, math.sqrt(2. / n))
107
            elif isinstance(m, nn.BatchNorm2d):
108
                m.weight.data.fill_(1)
109
                m.bias.data.zero_()
110

111
    def _make_layer(self, block, planes, blocks, stride=1):
112
        downsample = None
113
        if stride != 1 or self.inplanes != planes * block.expansion:
114
            downsample = nn.Sequential(
115
                nn.Conv2d(self.inplanes, planes * block.expansion,
116
                          kernel_size=1, stride=stride, bias=False),
117
                nn.BatchNorm2d(planes * block.expansion),
118
            )
119

120
        layers = []
121
        layers.append(block(self.inplanes, planes, stride, downsample))
122
        self.inplanes = planes * block.expansion
123
        for i in range(1, blocks):
124
            layers.append(block(self.inplanes, planes))
125

126
        return nn.Sequential(*layers)
127

128
    def forward(self, x):
129
        x0 = self.conv0(x)
130
        x0 = self.bn0(x0)
131
        x0 = self.relu0(x0)
132
        
133
        x1 = self.conv1(x0)
134
        x1 = self.bn1(x1)
135
        x1 = self.relu(x1)
136

137
        x2 = self.layer1(x1)
138
        x3 = self.layer2(x2)
139
        x4 = self.layer3(x3)
140
        x5 = self.layer4(x4)
141
        x6 = self.layer5(x5)
142
        x7 = self.layer6(x6)
143

144
        return x7, x6, x5, x4, x3, x2, x1, x0
145

146
    
147
def MLAttrEncoderResnet(**kwargs):
148
    model = ResNet(Bottleneck, [2, 2, 2, 2, 2, 2], **kwargs)
149
    return model
150