-
Notifications
You must be signed in to change notification settings - Fork 16
/
Copy pathResNeXt3D.py
268 lines (210 loc) · 8.36 KB
/
ResNeXt3D.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
from functools import partial
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.autograd import Variable
def conv3x3x3(in_planes, out_planes, stride=1):
"""3x3x3 convolution with padding."""
return nn.Conv3d(
in_planes,
out_planes,
kernel_size=3,
stride=stride,
padding=1,
bias=False)
def downsample_basic_block(x, planes, stride):
out = F.avg_pool3d(x, kernel_size=1, stride=stride)
zero_pads = torch.Tensor(
out.size(0), planes - out.size(1), out.size(2), out.size(3),
out.size(4)).zero_()
if isinstance(out.data, torch.cuda.FloatTensor):
zero_pads = zero_pads.cuda()
out = Variable(torch.cat([out.data, zero_pads], dim=1))
return out
class ResNeXtBottleneck(nn.Module):
expansion = 2
def __init__(self, inplanes, planes, cardinality, stride=1,
downsample=None):
super(ResNeXtBottleneck, self).__init__()
mid_planes = cardinality * int(planes / 32)
self.conv1 = nn.Conv3d(inplanes, mid_planes, kernel_size=1, bias=False)
self.gn1 = nn.GroupNorm(32, mid_planes)
# self.bn1 = nn.BatchNorm3d(mid_planes)
self.conv2 = nn.Conv3d(
mid_planes,
mid_planes,
kernel_size=3,
stride=stride,
padding=1,
groups=cardinality,
bias=False)
self.gn2 = nn.GroupNorm(32, mid_planes)
# self.bn2 = nn.BatchNorm3d(mid_planes)
self.conv3 = nn.Conv3d(
mid_planes, planes * self.expansion, kernel_size=1, bias=False)
# self.bn3 = nn.BatchNorm3d(planes * self.expansion)
self.gn3 = nn.GroupNorm(32, planes * self.expansion)
self.relu = nn.PReLU()
self.downsample = downsample
self.stride = stride
def forward(self, x):
residual = x
out = self.conv1(x)
out = self.gn1(out)
out = self.relu(out)
out = self.conv2(out)
out = self.gn2(out)
out = self.relu(out)
out = self.conv3(out)
out = self.gn3(out)
if self.downsample is not None:
residual = self.downsample(x)
out += residual
out = self.relu(out)
return out
class ResNeXtDilatedBottleneck(nn.Module):
expansion = 2
def __init__(self, inplanes, planes, cardinality, stride=1,
downsample=None):
super(ResNeXtDilatedBottleneck, self).__init__()
mid_planes = cardinality * int(planes / 32)
self.conv1 = nn.Conv3d(inplanes, mid_planes, kernel_size=1, bias=False)
# self.bn1 = nn.BatchNorm3d(mid_planes)
self.gn1 = nn.GroupNorm(32, mid_planes)
self.conv2 = nn.Conv3d(
mid_planes,
mid_planes,
kernel_size=3,
stride=stride,
padding=2,
dilation=2,
groups=cardinality,
bias=False)
# self.bn2 = nn.BatchNorm3d(mid_planes)
self.gn2 = nn.GroupNorm(32, mid_planes)
self.conv3 = nn.Conv3d(
mid_planes, planes * self.expansion, kernel_size=1, bias=False)
# self.bn3 = nn.BatchNorm3d(planes * self.expansion)
self.gn3 = nn.GroupNorm(32, planes * self.expansion)
self.relu = nn.PReLU()
self.downsample = downsample
self.stride = stride
def forward(self, x):
residual = x
out = self.conv1(x)
out = self.gn1(out)
out = self.relu(out)
out = self.conv2(out)
out = self.gn2(out)
out = self.relu(out)
out = self.conv3(out)
out = self.gn3(out)
if self.downsample is not None:
residual = self.downsample(x)
out += residual
out = self.relu(out)
return out
class ResNeXt3D(nn.Module):
def __init__(self, block, layers, shortcut_type='B', cardinality=32, num_classes=400):
self.inplanes = 64
super(ResNeXt3D, self).__init__()
self.conv1 = nn.Conv3d(1, 64, kernel_size=7, stride=(1, 2, 2), padding=(3, 3, 3), bias=False)
# self.bn1 = nn.BatchNorm3d(64)
self.gn1 = nn.GroupNorm(32, 64)
self.relu = nn.PReLU()
self.maxpool = nn.MaxPool3d(kernel_size=(3, 3, 3), stride=2, padding=1)
self.layer1 = self._make_layer(block, 128, layers[0], shortcut_type, cardinality)
self.layer2 = self._make_layer(block, 256, layers[1], shortcut_type, cardinality, stride=(1, 2, 2))
self.layer3 = self._make_layer(ResNeXtDilatedBottleneck, 512, layers[2], shortcut_type, cardinality, stride=1)
self.layer4 = self._make_layer(ResNeXtDilatedBottleneck, 1024, layers[3], shortcut_type, cardinality, stride=1)
self.avgpool = nn.AdaptiveAvgPool3d(1)
self.fc = nn.Linear(cardinality * 32 * block.expansion, num_classes)
for m in self.modules():
if isinstance(m, nn.Conv3d):
m.weight = nn.init.kaiming_normal(m.weight, mode='fan_out')
elif isinstance(m, nn.BatchNorm3d):
m.weight.data.fill_(1)
m.bias.data.zero_()
def _make_layer(self, block, planes, blocks, shortcut_type, cardinality, stride=1):
downsample = None
if stride != 1 or self.inplanes != planes * block.expansion:
if shortcut_type == 'A':
downsample = partial(
downsample_basic_block,
planes=planes * block.expansion,
stride=stride)
else:
downsample = nn.Sequential(
nn.Conv3d(
self.inplanes,
planes * block.expansion,
kernel_size=1,
stride=stride,
bias=False),
# nn.BatchNorm3d(planes * block.expansion)
nn.GroupNorm(32, planes * block.expansion),
)
layers = []
layers.append(
block(self.inplanes, planes, cardinality, stride, downsample))
self.inplanes = planes * block.expansion
for _ in range(1, blocks):
layers.append(block(self.inplanes, planes, cardinality))
return nn.Sequential(*layers)
def forward(self, x):
x = self.conv1(x)
x = self.gn1(x)
x = self.relu(x)
x = self.maxpool(x)
x = self.layer1(x)
x = self.layer2(x)
x = self.layer3(x)
x = self.layer4(x)
x = self.avgpool(x)
x = x.view(x.size(0), -1)
x = self.fc(x)
return x
def get_fine_tuning_parameters(model, ft_begin_index):
if ft_begin_index == 0:
return model.parameters()
ft_module_names = []
for i in range(ft_begin_index, 5):
ft_module_names.append('layer{}'.format(i))
ft_module_names.append('fc')
parameters = []
for k, v in model.named_parameters():
for ft_module in ft_module_names:
if ft_module in k:
parameters.append({'params': v})
break
else:
parameters.append({'params': v, 'lr': 0.0})
return parameters
def resnext3d10(**kwargs):
"""Constructs a ResNeXt3D-10 model."""
model = ResNeXt3D(ResNeXtBottleneck, [1, 1, 1, 1], **kwargs)
return model
def resnext3d18(**kwargs):
"""Constructs a ResNeXt3D-18 model."""
model = ResNeXt3D(ResNeXtBottleneck, [2, 2, 2, 2], **kwargs)
return model
def resnext3d34(**kwargs):
"""Constructs a ResNeXt3D-34 model."""
model = ResNeXt3D(ResNeXtBottleneck, [3, 4, 6, 3], **kwargs)
return model
def resnext3d50(**kwargs):
"""Constructs a ResNeXt3D-50 model."""
model = ResNeXt3D(ResNeXtBottleneck, [3, 4, 6, 3], **kwargs)
return model
def resnext3d101(**kwargs):
"""Constructs a ResNeXt3D-101 model."""
model = ResNeXt3D(ResNeXtBottleneck, [3, 4, 23, 3], **kwargs)
return model
def resnext3d152(**kwargs):
"""Constructs a ResNeXt3D-152 model."""
model = ResNeXt3D(ResNeXtBottleneck, [3, 8, 36, 3], **kwargs)
return model
def resnext3d200(**kwargs):
"""Constructs a ResNeXt3D-200 model."""
model = ResNeXt3D(ResNeXtBottleneck, [3, 24, 36, 3], **kwargs)
return model