-
Notifications
You must be signed in to change notification settings - Fork 607
/
Copy pathresnet.py
291 lines (218 loc) · 9.76 KB
/
resnet.py
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
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
import math
import torch.nn as nn
from collections import OrderedDict
import torch.utils.model_zoo as model_zoo
from torchvision.models.resnet import model_urls
from .base import Backbone
def conv3x3(in_planes, out_planes, stride=1, dilation=1):
"""3x3 convolution with padding"""
return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,
padding=dilation, bias=False, dilation=dilation)
class BasicBlock(nn.Module):
expansion = 1
def __init__(self, inplanes, planes, stride=1, downsample=None, dilation=1, use_bn=True):
super(BasicBlock, self).__init__()
self.use_bn = use_bn
self.conv1 = conv3x3(inplanes, planes, stride, dilation=dilation)
if use_bn:
self.bn1 = nn.BatchNorm2d(planes)
self.relu = nn.ReLU(inplace=True)
self.conv2 = conv3x3(planes, planes, dilation=dilation)
if use_bn:
self.bn2 = nn.BatchNorm2d(planes)
self.downsample = downsample
self.stride = stride
def forward(self, x):
residual = x
out = self.conv1(x)
if self.use_bn:
out = self.bn1(out)
out = self.relu(out)
out = self.conv2(out)
if self.use_bn:
out = self.bn2(out)
if self.downsample is not None:
residual = self.downsample(x)
out += residual
out = self.relu(out)
return out
class Bottleneck(nn.Module):
expansion = 4
def __init__(self, inplanes, planes, stride=1, downsample=None, dilation=1):
super(Bottleneck, self).__init__()
self.conv1 = nn.Conv2d(inplanes, planes, kernel_size=1, bias=False)
self.bn1 = nn.BatchNorm2d(planes)
self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=stride,
padding=dilation, bias=False, dilation=dilation)
self.bn2 = nn.BatchNorm2d(planes)
self.conv3 = nn.Conv2d(planes, planes * 4, kernel_size=1, bias=False)
self.bn3 = nn.BatchNorm2d(planes * 4)
self.relu = nn.ReLU(inplace=True)
self.downsample = downsample
self.stride = stride
def forward(self, x):
residual = 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)
if self.downsample is not None:
residual = self.downsample(x)
out += residual
out = self.relu(out)
return out
class ResNet(Backbone):
""" ResNet network module. Allows extracting specific feature blocks."""
def __init__(self, block, layers, output_layers, num_classes=1000, inplanes=64, dilation_factor=1, frozen_layers=()):
self.inplanes = inplanes
super(ResNet, self).__init__(frozen_layers=frozen_layers)
self.output_layers = output_layers
self.conv1 = nn.Conv2d(3, inplanes, kernel_size=7, stride=2, padding=3,
bias=False)
self.bn1 = nn.BatchNorm2d(inplanes)
self.relu = nn.ReLU(inplace=True)
self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
stride = [1 + (dilation_factor < l) for l in (8, 4, 2)]
self.layer1 = self._make_layer(block, inplanes, layers[0], dilation=max(dilation_factor//8, 1))
self.layer2 = self._make_layer(block, inplanes*2, layers[1], stride=stride[0], dilation=max(dilation_factor//4, 1))
self.layer3 = self._make_layer(block, inplanes*4, layers[2], stride=stride[1], dilation=max(dilation_factor//2, 1))
self.layer4 = self._make_layer(block, inplanes*8, layers[3], stride=stride[2], dilation=dilation_factor)
out_feature_strides = {'conv1': 4, 'layer1': 4, 'layer2': 4*stride[0], 'layer3': 4*stride[0]*stride[1],
'layer4': 4*stride[0]*stride[1]*stride[2]}
# TODO better way?
if isinstance(self.layer1[0], BasicBlock):
out_feature_channels = {'conv1': inplanes, 'layer1': inplanes, 'layer2': inplanes*2, 'layer3': inplanes*4,
'layer4': inplanes*8}
elif isinstance(self.layer1[0], Bottleneck):
base_num_channels = 4 * inplanes
out_feature_channels = {'conv1': inplanes, 'layer1': base_num_channels, 'layer2': base_num_channels * 2,
'layer3': base_num_channels * 4, 'layer4': base_num_channels * 8}
else:
raise Exception('block not supported')
self._out_feature_strides = out_feature_strides
self._out_feature_channels = out_feature_channels
# self.avgpool = nn.AvgPool2d(7, stride=1)
self.avgpool = nn.AdaptiveAvgPool2d((1,1))
self.fc = nn.Linear(inplanes*8 * block.expansion, num_classes)
for m in self.modules():
if isinstance(m, nn.Conv2d):
n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
m.weight.data.normal_(0, math.sqrt(2. / n))
elif isinstance(m, nn.BatchNorm2d):
m.weight.data.fill_(1)
m.bias.data.zero_()
def out_feature_strides(self, layer=None):
if layer is None:
return self._out_feature_strides
else:
return self._out_feature_strides[layer]
def out_feature_channels(self, layer=None):
if layer is None:
return self._out_feature_channels
else:
return self._out_feature_channels[layer]
def _make_layer(self, block, planes, blocks, stride=1, dilation=1):
downsample = None
if stride != 1 or self.inplanes != planes * block.expansion:
downsample = nn.Sequential(
nn.Conv2d(self.inplanes, planes * block.expansion,
kernel_size=1, stride=stride, bias=False),
nn.BatchNorm2d(planes * block.expansion),
)
layers = []
layers.append(block(self.inplanes, planes, stride, downsample, dilation=dilation))
self.inplanes = planes * block.expansion
for i in range(1, blocks):
layers.append(block(self.inplanes, planes))
return nn.Sequential(*layers)
def _add_output_and_check(self, name, x, outputs, output_layers):
if name in output_layers:
outputs[name] = x
return len(output_layers) == len(outputs)
def forward(self, x, output_layers=None):
""" Forward pass with input x. The output_layers specify the feature blocks which must be returned """
outputs = OrderedDict()
if output_layers is None:
output_layers = self.output_layers
x = self.conv1(x)
x = self.bn1(x)
x = self.relu(x)
if self._add_output_and_check('conv1', x, outputs, output_layers):
return outputs
x = self.maxpool(x)
x = self.layer1(x)
if self._add_output_and_check('layer1', x, outputs, output_layers):
return outputs
x = self.layer2(x)
if self._add_output_and_check('layer2', x, outputs, output_layers):
return outputs
x = self.layer3(x)
if self._add_output_and_check('layer3', x, outputs, output_layers):
return outputs
x = self.layer4(x)
if self._add_output_and_check('layer4', x, outputs, output_layers):
return outputs
x = self.avgpool(x)
x = x.view(x.size(0), -1)
x = self.fc(x)
if self._add_output_and_check('fc', x, outputs, output_layers):
return outputs
if len(output_layers) == 1 and output_layers[0] == 'default':
return x
raise ValueError('output_layer is wrong.')
def resnet_baby(output_layers=None, pretrained=False, inplanes=16, **kwargs):
"""Constructs a ResNet-18 model.
"""
if output_layers is None:
output_layers = ['default']
else:
for l in output_layers:
if l not in ['conv1', 'layer1', 'layer2', 'layer3', 'layer4', 'fc']:
raise ValueError('Unknown layer: {}'.format(l))
model = ResNet(BasicBlock, [2, 2, 2, 2], output_layers, inplanes=inplanes, **kwargs)
if pretrained:
raise NotImplementedError
return model
def resnet18(output_layers=None, pretrained=False, **kwargs):
"""Constructs a ResNet-18 model.
"""
if output_layers is None:
output_layers = ['default']
else:
for l in output_layers:
if l not in ['conv1', 'layer1', 'layer2', 'layer3', 'layer4', 'fc']:
raise ValueError('Unknown layer: {}'.format(l))
model = ResNet(BasicBlock, [2, 2, 2, 2], output_layers, **kwargs)
if pretrained:
model.load_state_dict(model_zoo.load_url(model_urls['resnet18']))
return model
def resnet50(output_layers=None, pretrained=False, **kwargs):
"""Constructs a ResNet-50 model.
"""
if output_layers is None:
output_layers = ['default']
else:
for l in output_layers:
if l not in ['conv1', 'layer1', 'layer2', 'layer3', 'layer4', 'fc']:
raise ValueError('Unknown layer: {}'.format(l))
model = ResNet(Bottleneck, [3, 4, 6, 3], output_layers, **kwargs)
if pretrained:
model.load_state_dict(model_zoo.load_url(model_urls['resnet50']))
return model
def resnet101(output_layers=None, pretrained=False, **kwargs):
"""Constructs a ResNet-101 model.
"""
if output_layers is None:
output_layers = ['default']
else:
for l in output_layers:
if l not in ['conv1', 'layer1', 'layer2', 'layer3', 'layer4', 'fc']:
raise ValueError('Unknown layer: {}'.format(l))
model = ResNet(Bottleneck, [3, 4, 23, 3], output_layers, **kwargs)
if pretrained:
model.load_state_dict(model_zoo.load_url(model_urls['resnet101']))
return model