-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathmodel.py
533 lines (477 loc) · 25.5 KB
/
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
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
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
import torch
import torch.nn as nn
# from torch.nn.utils import spectral_norm
from spectral_normalization import SpectralNorm as spectral_norm
def initialize_weights(net):
for m in net.modules():
if isinstance(m, torch.nn.Conv3d):
m.weight.data.normal_(0, 0.02)
# m.bias.data.zero_()
elif isinstance(m, torch.nn.ConvTranspose3d):
m.weight.data.normal_(0, 0.02)
# m.bias.data.zero_()
elif isinstance(m, torch.nn.Conv2d):
m.weight.data.normal_(0, 0.02)
# m.bias.data.zero_()
elif isinstance(m, torch.nn.ConvTranspose2d):
m.weight.data.normal_(0, 0.02)
# m.bias.data.zero_()
elif isinstance(m, torch.nn.Linear):
m.weight.data.normal_(0, 0.02)
# m.bias.data.zero_()
def print_network(net):
num_params = 0
for param in net.parameters():
num_params += param.numel()
print(net)
print('Total number of parameters: %d' % num_params)
class GLU(torch.nn.Module):
def __init__(self):
super(GLU, self).__init__()
def forward(self, x):
nc = x.size(1)
assert nc % 2 == 0, 'channels dont divide 2!'
nc = int(nc/2)
return x[:, :nc] * torch.nn.functional.sigmoid(x[:, nc:])
class CA_NET(torch.nn.Module):
# some code is modified from vae examples
# (https://github.com/pytorch/examples/blob/master/vae/main.py)
def __init__(self, args):
super(CA_NET, self).__init__()
self.args = args
self.t_dim = args.hidden_size
self.c_dim = 100
self.fc = torch.nn.Linear(self.t_dim, self.c_dim * 4, bias=True)
self.relu = GLU()
def encode(self, text_embedding):
x = self.relu(self.fc(text_embedding))
mu = x[:, :self.c_dim]
logvar = x[:, self.c_dim:]
return mu, logvar
def reparametrize(self, mu, logvar):
std = logvar.mul(0.5).exp_()
if self.args.cuda:
eps = torch.cuda.FloatTensor(std.size()).normal_()
else:
eps = torch.FloatTensor(std.size()).normal_()
# eps = Variable(eps)
return eps.mul(std).add_(mu)
def forward(self, text_embedding):
mu, logvar = self.encode(text_embedding)
c_code = self.reparametrize(mu, logvar)
return c_code, mu, logvar
class GlobalAttentionGeneral(nn.Module):
def __init__(self, idf, cdf):
super(GlobalAttentionGeneral, self).__init__()
# self.conv_context = conv1x1(cdf, idf)
self.conv_context = nn.Conv2d(cdf, idf, 1, 1, 0, bias=False)
self.sm = nn.Softmax()
self.mask = None
def applyMask(self, mask):
self.mask = mask # batch x sourceL
def forward(self, input, context):
"""
input: batch x channels x queryL x ih x iw (idf=ihxiwxchannels)
context: batch x cdf x sourceL
"""
cln, ih, iw = input.size(1), input.size(3), input.size(4)
# queryL = ih * iw * cln
idf = ih * iw * cln
queryL = input.size(2)
batch_size, sourceL = context.size(0), context.size(2)
# --> batch x queryL x idf
target = input.permute(0, 2, 1, 3, 4).contiguous()
targetT = target.view(batch_size, -1, idf)
# targetT = torch.transpose(target, 1, 2).contiguous()
# batch x cdf x sourceL --> batch x cdf x sourceL x 1
sourceT = context.unsqueeze(3)
# --> batch x idf x sourceL
sourceT = self.conv_context(sourceT).squeeze(3)
# Get attention
# (batch x queryL x idf)(batch x idf x sourceL)
# -->batch x queryL x sourceL
attn = torch.bmm(targetT, sourceT)
# --> batch*queryL x sourceL
attn = attn.view(batch_size*queryL, sourceL)
if self.mask is not None:
# batch_size x sourceL --> batch_size*queryL x sourceL
mask = self.mask.repeat(queryL, 1)
attn.data.masked_fill_(mask.data, -float('inf'))
attn = self.sm(attn) # Eq. (2)
# --> batch x queryL x sourceL
attn = attn.view(batch_size, queryL, sourceL)
# --> batch x sourceL x queryL
attn = torch.transpose(attn, 1, 2).contiguous()
# (batch x idf x sourceL)(batch x sourceL x queryL)
# --> batch x idf x queryL
weightedContext = torch.bmm(sourceT, attn)
# weightedContext = weightedContext.view(batch_size, -1, ih, iw)
# attn = attn.view(batch_size, -1, ih, iw)
weightedContext = weightedContext.view(batch_size, -1, cln, ih, iw)
weightedContext = weightedContext.permute(0, 2, 1, 3, 4).contiguous()
# attn = attn.view(batch_size, -1, cln, ih, iw)
# attn = attn.permute(0, 2, 1, 3, 4).contiguous()
return weightedContext, attn
# ========================== #
# Contain the Linear layer #
# ========================== #
class _G(torch.nn.Module):
def __init__(self, args):
super(_G, self).__init__()
self.args = args
self.ca_net = CA_NET(self.args)
self.fc = torch.nn.Sequential(
# torch.nn.Linear(self.args.z_size+self.args.hidden_size, 1024, bias=False), # 64x(100+256)->64x1024
torch.nn.Linear(self.args.z_size+100, 1024, bias=False), # 64x(100+256)->64x1024
torch.nn.BatchNorm1d(1024),
torch.nn.ReLU(inplace=True),
torch.nn.Linear(1024, 512 * 2 * (self.args.imageSize // 8) * (self.args.imageSize // 8), bias=False), # 64x1024->64x(512x2x8x8)
torch.nn.BatchNorm1d(512 * 2 * (self.args.imageSize // 8) * (self.args.imageSize // 8)),
torch.nn.ReLU(inplace=True),
)
self.deconv = torch.nn.Sequential(
torch.nn.ConvTranspose3d(512, 256, 4, 2, 1, bias=False), # 64x512x2x8x8->64x256x4x16x16
torch.nn.BatchNorm3d(256),
torch.nn.ReLU(inplace=True),
torch.nn.ConvTranspose3d(256, 128, 4, 2, 1, bias=False), # 64x256x4x16x16->64x128x8x32x32
torch.nn.BatchNorm3d(128),
torch.nn.ReLU(inplace=True),
torch.nn.ConvTranspose3d(128, 64, 4, 2, 1, bias=False), # 64x128x8x32x32->64x64x16x64x64
torch.nn.BatchNorm3d(64),
torch.nn.ReLU(inplace=True),
# torch.nn.ConvTranspose3d(64, 1, 3, 1, 1, bias=False), # 64x64x16x64x64->64x1x16x64x64
torch.nn.ConvTranspose3d(64, args.input_channels, 3, 1, 1, bias=False), # 64x64x16x64x64->64x1x16x64x64
torch.nn.Tanh(),
)
if self.args.init:
print("Initialize _G")
initialize_weights(self)
def forward(self, x, sent_emb):
c_code, mu, logvar = self.ca_net(sent_emb)
x_c_code = torch.cat((x, c_code), 1)
# print(x_c_code.size()) # torch.Size([64, 100+256])
out = self.fc(x_c_code)
# print(out.size()) # torch.Size([64, (512x2x8x8)])
out = out.view(-1, 512, 2, (self.args.imageSize // 8), (self.args.imageSize // 8))
# print(out.size()) # torch.Size([64, 512, 2, 8, 8])
out = self.deconv(out)
# print(out.size()) # torch.Size([64, 1, 16, 64, 64])
return out, mu, logvar
# # ============================= #
# # Not Contain the Linear layer #
# # ============================= #
# class _G(torch.nn.Module):
# def __init__(self, args):
# super(_G, self).__init__()
# self.args = args
# self.ca_net = CA_NET(self.args)
# self.deconv = torch.nn.Sequential(
# torch.nn.ConvTranspose3d(self.args.z_size+100, 512,
# (2, self.args.imageSize // 8, self.args.imageSize // 8), 1, 0, bias=False),
# torch.nn.BatchNorm3d(512),
# torch.nn.ReLU(inplace=True),
# torch.nn.ConvTranspose3d(512, 256, 4, 2, 1, bias=False), # 64x512x2x8x8->64x256x4x16x16
# torch.nn.BatchNorm3d(256),
# torch.nn.ReLU(inplace=True),
# torch.nn.ConvTranspose3d(256, 128, 4, 2, 1, bias=False), # 64x256x4x16x16->64x128x8x32x32
# torch.nn.BatchNorm3d(128),
# torch.nn.ReLU(inplace=True),
# torch.nn.ConvTranspose3d(128, 64, 4, 2, 1, bias=False), # 64x128x8x32x32->64x64x16x64x64
# torch.nn.BatchNorm3d(64),
# torch.nn.ReLU(inplace=True),
# # torch.nn.ConvTranspose3d(64, 1, 3, 1, 1, bias=False), # 64x64x16x64x64->64x1x16x64x64
# torch.nn.ConvTranspose3d(64, args.input_channels, 1, 1, 0, bias=False), # 64x64x16x64x64->64x1x16x64x64
# torch.nn.Tanh(),
# )
# if self.args.init:
# print("Initialize _G")
# initialize_weights(self)
# def forward(self, x, sent_emb):
# c_code, mu, logvar = self.ca_net(sent_emb)
# x_c_code = torch.cat((x, c_code), 1)
# # print(x_c_code.size()) # torch.Size([64, 100+256])
# x_c_code = x_c_code.view(-1, self.args.z_size+100, 1, 1, 1)
# out = self.deconv(x_c_code)
# # print(out.size()) # torch.Size([64, 1, 16, 64, 64])
# return out, mu, logvar
# class _G(torch.nn.Module):
# def __init__(self, args):
# super(_G, self).__init__()
# self.args = args
# self.ca_net = CA_NET(self.args)
# self.fc = torch.nn.Sequential(
# # torch.nn.Linear(self.args.z_size+self.args.hidden_size, 1024, bias=False), # 64x(100+256)->64x1024
# torch.nn.Linear(self.args.z_size+100, 1024, bias=False), # 64x(100+256)->64x1024
# torch.nn.BatchNorm1d(1024),
# torch.nn.ReLU(inplace=True),
# torch.nn.Linear(1024, 512 * 2 * (self.args.imageSize // 8) * (self.args.imageSize // 8), bias=False), # 64x1024->64x(512x2x8x8)
# torch.nn.BatchNorm1d(512 * 2 * (self.args.imageSize // 8) * (self.args.imageSize // 8)),
# torch.nn.ReLU(inplace=True),
# )
# self.deconv = torch.nn.Sequential(
# torch.nn.ConvTranspose3d(512, 256, 4, 2, 1, bias=False), # 64x512x2x8x8->64x256x4x16x16
# torch.nn.BatchNorm3d(256),
# torch.nn.ReLU(inplace=True),
# torch.nn.ConvTranspose3d(256, 128, 4, 2, 1, bias=False), # 64x256x4x16x16->64x128x8x32x32
# torch.nn.BatchNorm3d(128),
# torch.nn.ReLU(inplace=True),
# torch.nn.ConvTranspose3d(128, 64, 4, 2, 1, bias=False), # 64x128x8x32x32->64x64x16x64x64
# torch.nn.BatchNorm3d(64),
# torch.nn.ReLU(inplace=True),
# # torch.nn.ConvTranspose3d(64, 1, 3, 1, 1, bias=False), # 64x64x16x64x64->64x1x16x64x64
# torch.nn.ConvTranspose3d(64, args.input_channels, 3, 1, 1, bias=False), # 64x64x16x64x64->64x1x16x64x64
# torch.nn.Tanh(),
# )
# # for attention
# self.att = GlobalAttentionGeneral(args.input_channels*args.imageSize*args.imageSize, args.hidden_size)
# self.deconv2 = torch.nn.Sequential(
# torch.nn.ConvTranspose3d(args.input_channels*2, args.input_channels, 1, 1, 0, bias=False), # 64x64x16x64x64->64x1x16x64x64
# torch.nn.BatchNorm3d(args.input_channels),
# torch.nn.ReLU(inplace=True),
# torch.nn.ConvTranspose3d(args.input_channels, args.input_channels, 1, 1, 0, bias=False), # 64x64x16x64x64->64x1x16x64x64
# torch.nn.Tanh(),
# )
# if self.args.init:
# print("Initialize _G")
# initialize_weights(self)
# def forward(self, z_code, sent_emb, word_embs, mask):
# c_code, mu, logvar = self.ca_net(sent_emb)
# z_c_code = torch.cat((z_code, c_code), 1)
# # print(x_c_code.size()) # torch.Size([64, 100+256])
# out = self.fc(z_c_code)
# # print(out.size()) # torch.Size([64, (512x2x8x8)])
# out = out.view(-1, 512, 2, (self.args.imageSize // 8), (self.args.imageSize // 8))
# # print(out.size()) # torch.Size([64, 512, 2, 8, 8])
# out = self.deconv(out)
# # print(out.size()) # torch.Size([64, 1, 16, 64, 64])
# # out = out.view(-1, self.args.input_channels, self.args.imageSize, self.args.imageSize)
# # # print(out.size()) # torch.Size([64*16, 1, 64, 64])
# self.att.applyMask(mask)
# out2, att = self.att(out, word_embs)
# out1_2 = torch.cat((out, out2), 1)
# out1_2 = self.deconv2(out1_2)
# return out1_2, mu, logvar
class _D(torch.nn.Module):
def __init__(self, args):
super(_D, self).__init__()
self.args = args
# if self.args.norm_D=='bn':
self.conv1 = torch.nn.Sequential(
# torch.nn.Conv3d(1, 64, 4, 2, 1, bias=False), # 64x1x16x64x64->64x64x8x32x32
torch.nn.Conv3d(args.input_channels, 64, 4, 2, 1, bias=False), # 64x1x16x64x64->64x64x8x32x32
torch.nn.BatchNorm3d(64),
torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
torch.nn.Conv3d(64, 128, 4, 2, 1, bias=False), # 64x64x8x32x32->64x128x4x16x16
torch.nn.BatchNorm3d(128),
torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
torch.nn.Conv3d(128, 256, 4, 2, 1, bias=False), # 64x128x4x16x16->64x256x2x8x8
torch.nn.BatchNorm3d(256),
torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
torch.nn.Conv3d(256, 512, 4, 2, 1, bias=False), # 64x256x2x8x8->64x512x1x4x4
torch.nn.BatchNorm3d(512),
torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
)
self.conv2 = torch.nn.Sequential(
torch.nn.Conv3d(512+self.args.hidden_size, 512, 3, 1, 1, bias=False), # 64x(512+256)x1x4x4->64x512x1x4x4
torch.nn.BatchNorm3d(512),
torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
)
self.fc = torch.nn.Sequential(
torch.nn.Linear(512 * 1 * (self.args.imageSize // 16) * (self.args.imageSize // 16), 1024, bias=False),
torch.nn.BatchNorm1d(1024),
torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
torch.nn.Linear(1024, 1, bias=False),
torch.nn.Sigmoid(),
)
# elif self.args.norm_D=='sn':
# self.conv1 = torch.nn.Sequential(
# # torch.nn.Conv3d(1, 64, 4, 2, 1, bias=False), # 64x1x16x64x64->64x64x8x32x32
# spectral_norm(torch.nn.Conv3d(args.input_channels, 64, 4, 2, 1, bias=False)), # 64x1x16x64x64->64x64x8x32x32
# # spectral_norm(64),
# torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
# spectral_norm(torch.nn.Conv3d(64, 128, 4, 2, 1, bias=False)), # 64x64x8x32x32->64x128x4x16x16
# # spectral_norm(128),
# torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
# spectral_norm(torch.nn.Conv3d(128, 256, 4, 2, 1, bias=False)), # 64x128x4x16x16->64x256x2x8x8
# # spectral_norm(256),
# torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
# spectral_norm(torch.nn.Conv3d(256, 512, 4, 2, 1, bias=False)), # 64x256x2x8x8->64x512x1x4x4
# # spectral_norm(512),
# torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
# )
# self.conv2 = torch.nn.Sequential(
# spectral_norm(torch.nn.Conv3d(512+self.args.hidden_size, 512, 3, 1, 1, bias=False)), # 64x(512+256)x1x4x4->64x512x1x4x4
# # spectral_norm(512),
# torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
# )
# self.fc = torch.nn.Sequential(
# spectral_norm(torch.nn.Linear(512 * 1 * (self.args.imageSize // 16) * (self.args.imageSize // 16), 1024, bias=False)),
# # spectral_norm(1024),
# torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
# torch.nn.Linear(1024, 1, bias=False),
# torch.nn.Sigmoid(),
# )
if self.args.init:
print("Initialize _D")
initialize_weights(self)
def forward(self, x, embedding):
# print(x.size()) # torch.Size([64, 16, 64, 64])
# print(embedding.size()) # torch.Size([64, 256])
# out = x.view(-1, 1, 16, self.args.imageSize, self.args.imageSize)
out = x.view(-1, self.args.input_channels, 16, self.args.imageSize, self.args.imageSize)
# print(out.size()) # torch.Size([64, 1, 16, 64, 64])
out = self.conv1(out)
# print(out.size()) # torch.Size([64, 512, 1, 4, 4])
embedding = embedding.unsqueeze(2).unsqueeze(3).unsqueeze(4).expand(
embedding.size()[0], embedding.size()[1], 1, out.size()[3], out.size()[4])
# print(embedding.size()) # torch.Size([64, 256]) -> torch.Size([64, 256, 1, 4, 4])
out = torch.cat((out, embedding), 1)
# print(out.size()) # torch.Size([64, 768, 1, 4, 4])
out = self.conv2(out)
# print(out.size()) # torch.Size([64, 512, 1, 4, 4])
out = out.view(-1, 512 * 1 * (self.args.imageSize // 16) * (self.args.imageSize // 16))
# print(out.size()) # torch.Size([64, 512x1x4x4])
out = self.fc(out)
# print(out.size()) # torch.Size([64, 1])
out = out.squeeze()
return out
class _D_frame_motion(torch.nn.Module):
def __init__(self, args):
super(_D_frame_motion, self).__init__()
self.args = args
# if self.args.norm_D=='bn':
self.conv1 = torch.nn.Sequential(
# torch.nn.Conv2d(1, 64, 4, 2, 1, bias=False), # (64x16)x1x64x64->(64x16)x64x32x32
torch.nn.Conv2d(args.input_channels, 64, 4, 2, 1, bias=False), # (64x16)x1x64x64->(64x16)x64x32x32
torch.nn.BatchNorm2d(64),
torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
torch.nn.Conv2d(64, 128, 4, 2, 1, bias=False), # (64x16)x64x32x32->(64x16)x128x16x16
torch.nn.BatchNorm2d(128),
torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
torch.nn.Conv2d(128, 256, 4, 2, 1, bias=False), # (64x16)x128x16x16->(64x16)x256x8x8
torch.nn.BatchNorm2d(256),
torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
torch.nn.Conv2d(256, 512, 4, 2, 1, bias=False), # (64x16)x256x8x8->(64x16)x512x4x4
torch.nn.BatchNorm2d(512),
torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
)
self.conv_frame = torch.nn.Sequential(
torch.nn.Conv2d(512+self.args.hidden_size, 512, 3, 1, 1, bias=False), # (64x16)x(512+256)x4x4->(64x16)x512x4x4
torch.nn.BatchNorm2d(512),
torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
)
self.fc_frame = torch.nn.Sequential(
torch.nn.Linear(512 * (self.args.imageSize // 16) * (self.args.imageSize // 16), 1024, bias=False),
torch.nn.BatchNorm1d(1024),
torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
torch.nn.Linear(1024, 1, bias=False),
torch.nn.Sigmoid(),
)
if self.args.A:
self.conv_motion = torch.nn.Sequential(
torch.nn.Conv2d(512+self.args.hidden_size, 512, 3, 1, 1, bias=False), # (64x16)x(512+256)x4x4->(64x16)x512x4x4
torch.nn.BatchNorm2d(512),
torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
)
self.fc_motion = torch.nn.Sequential(
torch.nn.Linear(512 * (self.args.imageSize // 16) * (self.args.imageSize // 16), 1024, bias=False),
torch.nn.BatchNorm1d(1024),
torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
torch.nn.Linear(1024, 1, bias=False),
torch.nn.Sigmoid(),
)
# elif self.args.norm_D=='sn':
# self.conv1 = torch.nn.Sequential(
# # torch.nn.Conv2d(1, 64, 4, 2, 1, bias=False), # (64x16)x1x64x64->(64x16)x64x32x32
# spectral_norm(torch.nn.Conv2d(args.input_channels, 64, 4, 2, 1, bias=False)), # (64x16)x1x64x64->(64x16)x64x32x32
# # spectral_norm(64),
# torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
# spectral_norm(torch.nn.Conv2d(64, 128, 4, 2, 1, bias=False)), # (64x16)x64x32x32->(64x16)x128x16x16
# # spectral_norm(128),
# torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
# spectral_norm(torch.nn.Conv2d(128, 256, 4, 2, 1, bias=False)), # (64x16)x128x16x16->(64x16)x256x8x8
# # spectral_norm(256),
# torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
# spectral_norm(torch.nn.Conv2d(256, 512, 4, 2, 1, bias=False)), # (64x16)x256x8x8->(64x16)x512x4x4
# # spectral_norm(512),
# torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
# )
# self.conv_frame = torch.nn.Sequential(
# spectral_norm(torch.nn.Conv2d(512+self.args.hidden_size, 512, 3, 1, 1, bias=False)), # (64x16)x(512+256)x4x4->(64x16)x512x4x4
# # spectral_norm(512),
# torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
# )
# self.fc_frame = torch.nn.Sequential(
# spectral_norm(torch.nn.Linear(512 * (self.args.imageSize // 16) * (self.args.imageSize // 16), 1024, bias=False)),
# # spectral_norm(1024),
# torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
# torch.nn.Linear(1024, 1, bias=False),
# torch.nn.Sigmoid(),
# )
# if self.args.A:
# self.conv_motion = torch.nn.Sequential(
# spectral_norm(torch.nn.Conv2d(512+self.args.hidden_size, 512, 3, 1, 1, bias=False)), # (64x16)x(512+256)x4x4->(64x16)x512x4x4
# # spectral_norm(512),
# torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
# )
# self.fc_motion = torch.nn.Sequential(
# spectral_norm(torch.nn.Linear(512 * (self.args.imageSize // 16) * (self.args.imageSize // 16), 1024, bias=False)),
# # spectral_norm(1024),
# torch.nn.LeakyReLU(self.args.leak_value, inplace=True),
# torch.nn.Linear(1024, 1, bias=False),
# torch.nn.Sigmoid(),
# )
if self.args.init:
print("Initialize _D_frame_motion")
initialize_weights(self)
def forward(self, x, embedding):
# print(x.size()) # torch.Size([64, 16, 64, 64])
# print(embedding.size()) # torch.Size([64, 256])
# out = x.view(-1, 1, self.args.imageSize, self.args.imageSize)
out = x.view(-1, self.args.input_channels, self.args.imageSize, self.args.imageSize)
# print(out.size()) # torch.Size([64x16=1024, 1, 64, 64])
out = self.conv1(out)
# print(out.size()) # torch.Size([64x16=1024, 512, 4, 4])
if self.args.A:
out_motion = out.view(x.size()[0], self.args.frame_num, out.size()[1], \
out.size()[2], out.size()[3])
# print(out_motion.size()) # torch.Size([64, 16, 512, 4, 4])
out_motion_pre = out_motion[:, :-1, :, :, :].clone()
out_motion_post = out_motion[:, 1:, :, :, :].clone()
# out_motion = (out_motion_post - out_motion_pre)**2
out_motion = out_motion_post - out_motion_pre
# print(out_motion.size()) # torch.Size([64, 15, 512, 4, 4])
out_motion = out_motion.view(x.size()[0]*(self.args.frame_num-1), \
out.size()[1], out.size()[2], out.size()[3])
# print(out_motion.size()) # torch.Size([64x15=960, 512, 4, 4])
embedding_frame = embedding.repeat(1, out.size()[0]/embedding.size()[0]).view(out.size()[0], -1)
# print(embedding_frame.size()) # torch.Size([64, 256]) -> torch.Size([1024, 256])
embedding_frame = embedding_frame.unsqueeze(2).unsqueeze(3).expand(
embedding_frame.size()[0], embedding_frame.size()[1], out.size()[2], out.size()[3])
# print(embedding_frame.size()) # torch.Size([1024, 256]) -> torch.Size([1024, 256, 4, 4])
out = torch.cat((out, embedding_frame), 1)
# print(out.size()) # torch.Size([1024, 768, 4, 4])
out = self.conv_frame(out)
# print(out.size()) # torch.Size([64x16=1024, 512, 4, 4])
out = out.view(-1, 512 * (self.args.imageSize // 16) * (self.args.imageSize // 16))
# print(out.size()) # torch.Size([1024, 512x4x4])
out = self.fc_frame(out)
# print(out.size()) # torch.Size([1024, 1])
out = out.squeeze()
if self.args.A:
embedding_motion = embedding.repeat(1, out_motion.size()[0]/embedding.size()[0]).view(out_motion.size()[0], -1)
# print(embedding_motion.size()) # torch.Size([64, 256]) -> torch.Size([960, 256])
embedding_motion = embedding_motion.unsqueeze(2).unsqueeze(3).expand(
embedding_motion.size()[0], embedding_motion.size()[1], out_motion.size()[2], out_motion.size()[3])
# print(embedding_motion.size()) # torch.Size([960, 256]) -> torch.Size([960, 256, 4, 4])
out_motion = torch.cat((out_motion, embedding_motion), 1)
# print(out_motion.size()) # torch.Size([960, 768, 4, 4])
out_motion = self.conv_motion(out_motion)
# print(out_motion.size()) # torch.Size([64x15=960, 512, 4, 4])
out_motion = out_motion.view(-1, 512 * (self.args.imageSize // 16) * (self.args.imageSize // 16))
# print(out_motion.size()) # torch.Size([960, 512x4x4])
out_motion = self.fc_motion(out_motion)
# print(out.size()) # torch.Size([960, 1])
out_motion = out_motion.squeeze()
return out, out_motion
return out