-
Notifications
You must be signed in to change notification settings - Fork 7
/
Dynamic_Model.py
241 lines (199 loc) · 7.69 KB
/
Dynamic_Model.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
# Dynamic_Model.py
import torch
import torch.nn as nn
class MDGAN_S1_G(nn.Module):
def __init__(self, ngf):
super(MDGAN_S1_G, self).__init__()
# input is 3 x 32 x 128 x 128 (duplicated by 3 x 1 x 128 x 128)
self.downConv1 = nn.Conv3d(3, ngf,(3, 4, 4), stride=(1, 2, 2), padding=(1, 1, 1),bias=False)
self.downConv2 = nn.Conv3d(ngf, ngf *2 ,(4, 4, 4), stride=(2, 2, 2), padding=(1, 1, 1),bias=False)
self.downConv3 = nn.Conv3d(ngf *2, ngf * 4, 4, 2, 1, bias=False)
self.downConv4 = nn.Conv3d(ngf * 4, ngf * 8, 4, 2, 1, bias=False)
self.downConv5 = nn.Conv3d(ngf * 8, ngf * 16, 4, 2, 1, bias=False)
self.downConv6 = nn.Conv3d(ngf * 16, ngf * 16, (2, 4, 4), stride=(1, 1, 1), padding=(0, 0, 0), bias=False)
# get
self.downBN2 = nn.BatchNorm3d(ngf * 2)
self.downBN3 = nn.BatchNorm3d(ngf * 4)
self.downBN4 = nn.BatchNorm3d(ngf * 8)
self.downBN5 = nn.BatchNorm3d(ngf * 16)
self.relu = nn.ReLU(inplace = True)
self.upConv1 = nn.ConvTranspose3d(ngf * 16, ngf * 16, (2,4,4), stride=(1, 1, 1), padding=(0, 0, 0), bias=False )
self.upConv2 = nn.ConvTranspose3d(ngf * 16, ngf * 8, 4, 2, 1, bias=False)
self.upConv3 = nn.ConvTranspose3d(ngf * 8, ngf * 4, 4, 2, 1, bias=False)
self.upConv4 = nn.ConvTranspose3d(ngf * 4, ngf * 2, 4, 2, 1, bias=False)
self.upConv5 = nn.ConvTranspose3d(ngf * 2, ngf * 1, (4,4,4), stride=(2, 2, 2), padding=(1, 1, 1), bias=False)
self.upConv6 = nn.ConvTranspose3d(ngf * 1, 3, (3,4,4), stride=(1, 2, 2), padding=(1, 1, 1), bias=False)
self.tanh = nn.Tanh()
self.upBN1 = nn.BatchNorm3d(ngf * 16)
self.upBN2 = nn.BatchNorm3d(ngf * 8)
self.upBN3 = nn.BatchNorm3d(ngf * 4)
self.upBN4 = nn.BatchNorm3d(ngf * 2)
self.upBN5 = nn.BatchNorm3d(ngf * 1)
self.lrelu = nn.LeakyReLU(0.2, inplace=True)
def forward(self, x):
downx1 = self.downConv1(x)
downx2 = self.downConv2(downx1)
downx2 = self.downBN2(downx2)
downx2 = self.lrelu(downx2)
downx3 = self.downConv3(downx2)
downx3 = self.downBN3(downx3)
downx3 = self.lrelu(downx3)
downx4 = self.downConv4(downx3)
downx4 = self.downBN4(downx4)
downx4 = self.lrelu(downx4)
downx5 = self.downConv5(downx4)
downx5 = self.downBN5(downx5)
downx5 = self.lrelu(downx5)
downx6 = self.downConv6(downx5)
upx1 = self.upConv1(downx6)
upx1 = self.upBN1(upx1)
upx1 = self.relu(upx1)
upx1 = downx5 + upx1
upx2 = self.upConv2(upx1)
upx2 = self.upBN2(upx2)
upx2 = self.relu(upx2)
upx2 = downx4 + upx2
upx3 = self.upConv3(upx2)
upx3 = self.upBN3(upx3)
upx3 = self.relu(upx3)
upx3 = downx3 + upx3
upx4 = self.upConv4(upx3)
upx4 = self.upBN4(upx4)
upx4 = self.relu(upx4)
upx4 = downx2 + upx4
upx5 = self.upConv5(upx4)
upx5 = self.upBN5(upx5)
upx5 = self.relu(upx5)
upx5 = downx1 + upx5
upx6 = self.upConv6(upx5)
upx6 = self.tanh(upx6)
return upx6
class MDGAN_S2_G(nn.Module):
def __init__(self, ngf):
super(MDGAN_S2_G, self).__init__()
# input is 3 x 32 x 128 x 128 (duplicated by 3 x 1 x 128 x 128)
self.downConv1 = nn.Conv3d(3, ngf,(3, 4, 4), stride=(1, 2, 2), padding=(1, 1, 1),bias=False)
self.downConv2 = nn.Conv3d(ngf, ngf *2 ,(4, 4, 4), stride=(2, 2, 2), padding=(1, 1, 1),bias=False)
self.downConv3 = nn.Conv3d(ngf *2, ngf * 4, 4, 2, 1, bias=False)
self.downConv4 = nn.Conv3d(ngf * 4, ngf * 8, 4, 2, 1, bias=False)
self.downConv5 = nn.Conv3d(ngf * 8, ngf * 16, 4, 2, 1, bias=False)
self.downConv6 = nn.Conv3d(ngf * 16, ngf * 16, (2, 4, 4), stride=(1, 1, 1), padding=(0, 0, 0), bias=False)
# get
self.downBN2 = nn.BatchNorm3d(ngf * 2)
self.downBN3 = nn.BatchNorm3d(ngf * 4)
self.downBN4 = nn.BatchNorm3d(ngf * 8)
self.downBN5 = nn.BatchNorm3d(ngf * 16)
self.relu = nn.ReLU(inplace = True)
self.upConv1 = nn.ConvTranspose3d(ngf * 16, ngf * 16, (2,4,4), stride=(1, 1, 1), padding=(0, 0, 0), bias=False )
self.upConv2 = nn.ConvTranspose3d(ngf * 16, ngf * 8, 4, 2, 1, bias=False)
self.upConv3 = nn.ConvTranspose3d(ngf * 8, ngf * 4, 4, 2, 1, bias=False)
self.upConv4 = nn.ConvTranspose3d(ngf * 4, ngf * 2, 4, 2, 1, bias=False)
self.upConv5 = nn.ConvTranspose3d(ngf * 2, ngf * 1, (4,4,4), stride=(2, 2, 2), padding=(1, 1, 1), bias=False)
self.upConv6 = nn.ConvTranspose3d(ngf * 1, 3, (3,4,4), stride=(1, 2, 2), padding=(1, 1, 1), bias=False)
self.tanh = nn.Tanh()
self.upBN1 = nn.BatchNorm3d(ngf * 16)
self.upBN2 = nn.BatchNorm3d(ngf * 8)
self.upBN3 = nn.BatchNorm3d(ngf * 4)
self.upBN4 = nn.BatchNorm3d(ngf * 2)
self.upBN5 = nn.BatchNorm3d(ngf * 1)
self.lrelu = nn.LeakyReLU(0.2, inplace=True)
def forward(self, x):
downx1 = self.downConv1(x)
downx2 = self.downConv2(downx1)
downx2 = self.downBN2(downx2)
downx2 = self.lrelu(downx2)
downx3 = self.downConv3(downx2)
downx3 = self.downBN3(downx3)
downx3 = self.lrelu(downx3)
downx4 = self.downConv4(downx3)
downx4 = self.downBN4(downx4)
downx4 = self.lrelu(downx4)
downx5 = self.downConv5(downx4)
downx5 = self.downBN5(downx5)
downx5 = self.lrelu(downx5)
downx6 = self.downConv6(downx5)
upx1 = self.upConv1(downx6)
upx1 = self.upBN1(upx1)
upx1 = self.relu(upx1)
upx1 = downx5 + upx1
upx2 = self.upConv2(upx1)
upx2 = self.upBN2(upx2)
upx2 = self.relu(upx2)
upx2 = downx4 + upx2
upx3 = self.upConv3(upx2)
upx3 = self.upBN3(upx3)
upx3 = self.relu(upx3)
upx3 = downx3 + upx3
upx4 = self.upConv4(upx3)
upx4 = self.upBN4(upx4)
upx4 = self.relu(upx4)
#upx4 = downx2 + upx4
upx5 = self.upConv5(upx4)
upx5 = self.upBN5(upx5)
upx5 = self.relu(upx5)
#upx5 = downx1 + upx5
upx6 = self.upConv6(upx5)
upx6 = self.tanh(upx6)
return upx6
class MDGAN_S2_D(nn.Module):
def __init__(self, ndf):
super(MDGAN_S2_D, self).__init__()
self.slice1 = nn.Sequential(
# input is 3 x 32 x 256 x 256
nn.Conv3d(3, ndf,(3, 4, 4), stride=(1, 2, 2), padding=(1, 1, 1),bias=False),
nn.LeakyReLU(0.2, inplace=True),
)
self.slice2 = nn.Sequential(
# ndf x 32 x 64 x 64
nn.Conv3d(ndf, ndf *2 ,(4, 4, 4), stride=(2, 2, 2), padding=(1, 1, 1),bias=False),
nn.BatchNorm3d(ndf * 2),
nn.LeakyReLU(0.2, inplace=True),
# (ndf*2) x 16 x 32 x 32
nn.Conv3d(ndf *2, ndf * 4, 4, 2, 1, bias=False),
nn.BatchNorm3d(ndf * 4),
nn.LeakyReLU(0.2, inplace=True),
)
self.slice3 = nn.Sequential(
# (ndf*4) x 8 x 16 x 16
nn.Conv3d(ndf * 4, ndf * 8, 4, 2, 1, bias=False),
nn.BatchNorm3d(ndf * 8),
nn.LeakyReLU(0.2, inplace=True),
# (ndf*8) x 4 x 8 x 8
nn.Conv3d(ndf * 8, ndf * 16, 4, 2, 1, bias=False),
nn.BatchNorm3d(ndf * 16),
nn.LeakyReLU(inplace=True),
)
self.slice4 = nn.Sequential(
# (ndf*16) x 2 x 4 x 4
nn.Conv3d(ndf * 16, 1, (2, 4, 4), stride=(1, 1, 1), padding=(0, 0, 0), bias=False),
nn.Sigmoid(),
)
def forward(self, x):
x1 = self.slice1(x)
x2 = self.slice2(x1)
x3 = self.slice3(x2)
x4 = self.slice4(x3)
return x4.view(-1, 1), [x2, x1]
class Sampler(object):
"""Base class for all Samplers.
Every Sampler subclass has to provide an __iter__ method, providing a way
to iterate over indices of dataset elements, and a __len__ method that
returns the length of the returned iterators.
"""
def __init__(self, data_source):
pass
def __iter__(self):
raise NotImplementedError
def __len__(self):
raise NotImplementedError
class SubsetRandomSampler(Sampler):
"""Samples elements randomly from a given list of indices, without replacement.
Arguments:
indices (list): a list of indices
"""
def __init__(self, indices):
self.indices = indices
def __iter__(self):
return (self.indices[i] for i in self.indices)
def __len__(self):
return len(self.indices)