resnet-pytorch/src/resnet.py at master · mbsariyildiz/resnet-pytorch · GitHub

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
import torch
import torch.nn as nn
import nn_ops

def _down_sample(x):
  return nn.functional.avg_pool2d(x, 2, 2)

def _increase_planes(x, n_out_planes):
  n_samples, n_planes, spatial_size = x.size()[:-1]
  x_zeros = torch.zeros(
    n_samples, n_out_planes - n_planes, spatial_size, spatial_size,
    dtype=x.dtype, device=x.device)
  return torch.cat([x, x_zeros], 1)

def _downsample_and_increase_planes(x, n_out_planes):
  x = _down_sample(x)
  x = _increase_planes(x, n_out_planes)
  return x

def identity_func(n_in_planes, n_out_planes, stride):
  identity = lambda x: x
  if stride == 2 and n_in_planes != n_out_planes:
    identity = lambda x: _downsample_and_increase_planes(x, n_out_planes)
  elif stride == 2:
    identity = _down_sample
  elif n_in_planes != n_out_planes:
    identity = lambda x: _increase_planes(x, n_out_planes)
  return identity

class BasicBlock(nn.Module):

  expansion = 1

  def __init__(self, n_in_planes, n_out_planes, stride=1):
    super().__init__()
    assert stride == 1 or stride == 2

    self.block = nn.Sequential(
      nn_ops.conv3x3(n_in_planes, n_out_planes, stride),
      nn.BatchNorm2d(n_out_planes),
      nn.ReLU(inplace=True),
      nn_ops.conv3x3(n_out_planes, n_out_planes),
      nn.BatchNorm2d(n_out_planes)
    )

    self.identity = identity_func(n_in_planes, n_out_planes, stride)

  def forward(self, x):
    out = self.block(x)
    identity = self.identity(x)

    out += identity
    out = nn.functional.relu(out)
    return out

class Bottleneck(nn.Module):

  expansion = 4

  def __init__(self, n_in_planes, n_out_planes, stride=1):
    super().__init__()

    self.conv1 = nn.Conv2d(n_in_planes, n_out_planes, kernel_size=1)
    self.bn1 = nn.BatchNorm2d(n_out_planes)

    self.conv2 = nn_ops.conv3x3(n_out_planes, n_out_planes, stride)
    self.bn2 = nn.BatchNorm2d(n_out_planes)

    self.conv3 = nn.Conv2d(n_out_planes, n_out_planes * 4, kernel_size=1)
    self.bn3 = nn.BatchNorm2d(n_out_planes * 4)

    self.relu = nn.ReLU(inplace=True)
    self.identity = identity_func(n_in_planes, n_out_planes * 4, stride)

  def forward(self, x):
    out = self.conv1(x)
    out = self.bn1(out)
    out = self.relu(out)

    out = self.conv2(out)
    out = self.bn2(out)
    out = self.relu(out)

    out = self.conv3(out)
    out = self.bn3(out)

    identity = self.identity(x)
    out += identity
    out = self.relu(out)

    return out

class ResNet(nn.Module):

  def __init__(self, block,
                     n_blocks,
                     n_output_planes,
                     n_classes):
    super(ResNet, self).__init__()
    assert len(n_blocks) == 4
    assert len(n_output_planes) == 4

    self.n_in_planes = n_output_planes[0]

    self.layer0 = nn.Sequential(
      nn_ops.conv3x3(3, self.n_in_planes),
      nn.BatchNorm2d(self.n_in_planes),
      nn.ReLU(inplace=True)
    )
    self.layer1 = self._make_layer(block, n_blocks[0], n_output_planes[0])
    self.layer2 = self._make_layer(block, n_blocks[1], n_output_planes[1], 2)
    self.layer3 = self._make_layer(block, n_blocks[2], n_output_planes[2], 2)
    self.layer4 = self._make_layer(block, n_blocks[3], n_output_planes[3], 2)
    self.fc = nn.Linear(n_output_planes[3] * block.expansion, n_classes, False)

    self.apply(nn_ops.variable_init)

  def _make_layer(self, block, n_blocks, n_out_planes, stride=1):
    layers = []
    layers.append(block(self.n_in_planes, n_out_planes, stride))
    self.n_in_planes = n_out_planes * block.expansion
    for i in range(1, n_blocks):
      layers.append(block(self.n_in_planes, n_out_planes))

    return nn.Sequential(*layers)

  def features(self, x):
    x = self.layer0(x)
    x = self.layer1(x)
    x = self.layer2(x)
    x = self.layer3(x)
    x = self.layer4(x)
    spatial_size = x.size(2)
    x = nn.functional.avg_pool2d(x, spatial_size, 1)
    x = x.view(x.size(0), -1)
    return x

  def forward(self, x):
    x = self.features(x)
    x = self.fc(x)
    return x

def ResNet10(**kwargs):
  return ResNet(BasicBlock, [1,1,1,1], **kwargs)

def ResNet18(**kwargs):
  return ResNet(BasicBlock, [2,2,2,2], **kwargs)

def ResNet34(**kwargs):
  return ResNet(BasicBlock, [3,4,6,3], **kwargs)

def ResNet50(**kwargs):
  return ResNet(Bottleneck, [3,4,6,3], **kwargs)