-
Notifications
You must be signed in to change notification settings - Fork 1
/
UGATIT.py
502 lines (438 loc) · 30 KB
/
UGATIT.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
#encoding=utf-8
import transforms
import dataset
from networks import *
from utils import *
import paddle.fluid as fluid
import glob,time
import paddle
from nn import BCEWithLogitsLoss
class UGATIT(object) :
def __init__(self, args):
self.light = args.light
if self.light :
self.model_name = 'UGATIT_light'
else :
self.model_name = 'UGATIT'
self.result_dir = args.result_dir
self.dataset = args.dataset
self.iteration = args.iteration
self.decay_flag = args.decay_flag
self.batch_size = args.batch_size
self.print_freq = args.print_freq
self.save_freq = args.save_freq
self.lr = args.lr
self.weight_decay = args.weight_decay
self.ch = args.ch
""" Weight """
self.adv_weight = args.adv_weight
self.cycle_weight = args.cycle_weight
self.identity_weight = args.identity_weight
self.cam_weight = args.cam_weight
""" Generator """
self.n_res = args.n_res
self.light_pool_size = args.light_pool_size
""" Discriminator """
self.n_dis = args.n_dis
self.img_size = args.img_size
self.img_ch = args.img_ch
self.device = fluid.CPUPlace() if "cpu" in args.device else fluid.CUDAPlace(0)
self.resume = args.resume
#self.logger = LogWriter("./log/train")
print()
print("##### Information #####")
print("# light : ", self.light)
print("# dataset : ", self.dataset)
print("# batch_size : ", self.batch_size)
print("# iteration per epoch : ", self.iteration)
print()
print("##### Generator #####")
print("# residual blocks : ", self.n_res)
print()
print("##### Discriminator #####")
print("# discriminator layer : ", self.n_dis)
print()
print("##### Weight #####")
print("# adv_weight : ", self.adv_weight)
print("# cycle_weight : ", self.cycle_weight)
print("# identity_weight : ", self.identity_weight)
print("# cam_weight : ", self.cam_weight)
##################################################################################
# Model
##################################################################################
def build_model(self):
""" DataLoader """
train_transform_A = [
transforms.RandomHorizontalFlip(),
transforms.Resize((self.img_size + 30, self.img_size+30)),
transforms.RandomCrop(self.img_size),
transforms.Normalize(mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5)),
transforms.ToCHWImage()
]
train_transform_B = [
transforms.RandomHorizontalFlip(),
#transforms.RandomRotate(),
transforms.Resize((self.img_size + 30, self.img_size + 30)),
transforms.RandomCrop(self.img_size),
transforms.Normalize(mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5)),
transforms.ToCHWImage()
]
test_transform = [
transforms.Resize((self.img_size, self.img_size)),
transforms.Normalize(mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5)),
transforms.ToCHWImage()
]
self.trainA_loader = paddle.batch(reader=paddle.reader.shuffle(reader=dataset.custom_reader(os.path.join('dataset', self.dataset, 'trainA'),
train_transform_A), buf_size=256), batch_size=self.batch_size,drop_last=False)
self.trainB_loader = paddle.batch(
reader=paddle.reader.shuffle(reader=dataset.custom_reader(os.path.join('dataset', self.dataset, 'trainB'),
train_transform_B), buf_size=256),batch_size=self.batch_size,drop_last=False)
self.testA_loader = paddle.batch(reader=paddle.reader.shuffle(dataset.custom_reader(os.path.join('dataset', self.dataset, 'testA'),
test_transform), buf_size=256), batch_size=self.batch_size,drop_last=False)
self.testB_loader = paddle.batch(reader = paddle.reader.shuffle(dataset.custom_reader(os.path.join('dataset', self.dataset, 'testB'),
test_transform), buf_size=256), batch_size=self.batch_size,drop_last=False)
with fluid.dygraph.guard(self.device):
""" Define Generator, Discriminator """
self.genA2B = ResnetGenerator(input_nc=3, output_nc=3, ngf=self.ch, n_blocks=self.n_res,
img_size=self.img_size, light=self.light)
self.genB2A = ResnetGenerator(input_nc=3, output_nc=3, ngf=self.ch, n_blocks=self.n_res,
img_size=self.img_size, light=self.light)
# global discriminator
self.disGA = Discriminator(input_nc=3, ndf=self.ch, n_layers=5)
self.disGB = Discriminator(input_nc=3, ndf=self.ch, n_layers=5)
# local discriminator
self.disLA = Discriminator(input_nc=3, ndf=self.ch, n_layers=3)
self.disLB = Discriminator(input_nc=3, ndf=self.ch, n_layers=3)
""" Define Loss """
self.L1_loss = fluid.dygraph.L1Loss()
self.MSE_loss = fluid.dygraph.MSELoss()
self.BCE_loss = BCEWithLogitsLoss()
self.l2_reg = fluid.regularizer.L2Decay(regularization_coeff=self.weight_decay)
""" Trainer """
boundaries = [self.iteration // 2]
values = [self.lr, self.lr - (self.lr / (self.iteration // 2))]
self.learning_rate = self.lr
if self.weight_decay:
self.learning_rate = fluid.dygraph.PiecewiseDecay(boundaries=boundaries, values=values,begin=0)
self.G_optim = fluid.optimizer.AdamOptimizer(
parameter_list=self.genA2B.parameters()+self.genB2A.parameters(),
learning_rate=self.learning_rate, beta1=0.5, beta2=0.999, regularization=self.l2_reg)
self.D_optim = fluid.optimizer.AdamOptimizer(
parameter_list=self.disGA.parameters()+self.disGB.parameters()+self.disLA.parameters()+self.disLB.parameters(),
learning_rate=self.learning_rate, beta1=0.5, beta2=0.999, regularization=self.l2_reg)
def train(self):
with fluid.dygraph.guard(self.device):
self.genA2B.train(), self.genB2A.train(), self.disGA.train(), self.disGB.train(), self.disLA.train(), self.disLB.train()
start_iter = 1
if self.resume:
model_list = glob.glob(os.path.join(self.result_dir,self.dataset,'*.*'))
if not len(model_list) == 0:
model_list.sort()
start_iter = int(model_list[-1].split('_')[-1].split('.')[0])
self.load(os.path.join(self.result_dir, self.dataset), start_iter)
print(" [*] Load SUCCESS")
# training loop
print('training start !')
start_time = time.time()
for step in range(start_iter, self.iteration + 1):
real_A_iter = next(self.trainA_loader())
real_B_iter = next(self.trainB_loader())
real_A, _ = real_A_iter[0][0], real_A_iter[0][1]
real_B, _ = real_B_iter[0][0], real_B_iter[0][1]
real_A = real_A[np.newaxis, :]
real_B = real_B[np.newaxis, :]
'''
论文中总共有4种loss:Adversarial loss,Cycle loss,Identity loss,CAM loss
其中，生成器用到了全部loss,判别器只用到了Adversarial loss和CAM loss
'''
real_A, real_B = fluid.dygraph.to_variable(real_A), fluid.dygraph.to_variable(real_B)
#set_requires_grad([self.genA2B, self.genB2A], False)
# Update D
self.D_optim.clear_gradients()
fake_A2B, _, _ = self.genA2B(real_A)
fake_B2A, _, _ = self.genB2A(real_B)
real_GA_logit, real_GA_cam_logit, _ = self.disGA(real_A)
real_LA_logit, real_LA_cam_logit, _ = self.disLA(real_A)
real_GB_logit, real_GB_cam_logit, _ = self.disGB(real_B)
real_LB_logit, real_LB_cam_logit, _ = self.disLB(real_B)
fake_GA_logit, fake_GA_cam_logit, _ = self.disGA(fake_B2A)
fake_LA_logit, fake_LA_cam_logit, _ = self.disLA(fake_B2A)
fake_GB_logit, fake_GB_cam_logit, _ = self.disGB(fake_A2B)
fake_LB_logit, fake_LB_cam_logit, _ = self.disLB(fake_A2B)
#判别器A的目的是让来自A的图片判断为真，生成的A的图片判断为假
#区分目标域真假图片，见论文公式(6)
D_ad_loss_GA = self.MSE_loss(real_GA_logit,
fluid.layers.ones_like(real_GA_logit)) + \
self.MSE_loss(fake_GA_logit,
fluid.layers.zeros_like(fake_GA_logit))
D_ad_cam_loss_GA = self.MSE_loss(real_GA_cam_logit,
fluid.layers.ones_like(real_GA_cam_logit)) + \
self.MSE_loss(fake_GA_cam_logit,
fluid.layers.zeros_like(fake_GA_cam_logit))
D_ad_loss_LA = self.MSE_loss(real_LA_logit,
fluid.layers.ones_like(real_LA_logit)) + \
self.MSE_loss(fake_LA_logit,
fluid.layers.zeros_like(fake_LA_logit))
D_ad_cam_loss_LA = self.MSE_loss(real_LA_cam_logit,
fluid.layers.ones_like(real_LA_cam_logit)) + \
self.MSE_loss(fake_LA_cam_logit,
fluid.layers.zeros_like(fake_LA_cam_logit))
# 判别器B的目的是让来自B的图片判断为真，生成的B的图片判断为假
D_ad_loss_GB = self.MSE_loss(real_GB_logit,
fluid.layers.ones_like(real_GB_logit)) + \
self.MSE_loss(fake_GB_logit,
fluid.layers.zeros_like(fake_GB_logit))
D_ad_cam_loss_GB = self.MSE_loss(real_GB_cam_logit,
fluid.layers.ones_like(real_GB_cam_logit)) + \
self.MSE_loss(fake_GB_cam_logit,
fluid.layers.zeros_like(fake_GB_cam_logit))
D_ad_loss_LB = self.MSE_loss(real_LB_logit,
fluid.layers.ones_like(real_LB_logit)) + \
self.MSE_loss(fake_LB_logit,
fluid.layers.zeros_like(fake_LB_logit))
D_ad_cam_loss_LB = self.MSE_loss(real_LB_cam_logit,
fluid.layers.ones_like(real_LB_cam_logit)) + \
self.MSE_loss(fake_LB_cam_logit,
fluid.layers.zeros_like(fake_LB_cam_logit))
D_loss_A = self.adv_weight * (D_ad_loss_GA + D_ad_cam_loss_GA + D_ad_loss_LA + D_ad_cam_loss_LA)
D_loss_B = self.adv_weight * (D_ad_loss_GB + D_ad_cam_loss_GB + D_ad_loss_LB + D_ad_cam_loss_LB)
Discriminator_loss = D_loss_A + D_loss_B
Discriminator_loss.backward()
self.D_optim.minimize(Discriminator_loss)
#set_requires_grad([self.genA2B, self.genB2A], True)
# Update G
self.G_optim.clear_gradients()
fake_A2B, fake_A2B_cam_logit, _ = self.genA2B(real_A)
fake_B2A, fake_B2A_cam_logit, _ = self.genB2A(real_B)
fake_A2B2A, _, _ = self.genB2A(fake_A2B) # 把用A域图片生成的B域图片再转到A域
fake_B2A2B, _, _ = self.genA2B(fake_B2A) # 把用B域图片生成的A域图片再转到B域
fake_A2A, fake_A2A_cam_logit, _ = self.genB2A(real_A) # 用A域图片生成A域图片
fake_B2B, fake_B2B_cam_logit, _ = self.genA2B(real_B) # 用B域图片生成B域图片
fake_GA_logit, fake_GA_cam_logit, _ = self.disGA(fake_B2A)
fake_LA_logit, fake_LA_cam_logit, _ = self.disLA(fake_B2A)
fake_GB_logit, fake_GB_cam_logit, _ = self.disGB(fake_A2B)
fake_LB_logit, fake_LB_cam_logit, _ = self.disLB(fake_A2B)
#生成器的作用就是以假乱真
G_ad_loss_GA = self.MSE_loss(fake_GA_logit,
fluid.layers.ones_like(fake_GA_logit))
G_ad_cam_loss_GA = self.MSE_loss(fake_GA_cam_logit,
fluid.layers.ones_like(fake_GA_cam_logit))
G_ad_loss_LA = self.MSE_loss(fake_LA_logit,
fluid.layers.ones_like(fake_LA_logit))
G_ad_cam_loss_LA = self.MSE_loss(fake_LA_cam_logit,
fluid.layers.ones_like(fake_LA_cam_logit))
G_ad_loss_GB = self.MSE_loss(fake_GB_logit,
fluid.layers.ones_like(fake_GB_logit))
G_ad_cam_loss_GB = self.MSE_loss(fake_GB_cam_logit,
fluid.layers.ones_like(fake_GB_cam_logit))
G_ad_loss_LB = self.MSE_loss(fake_LB_logit,
fluid.layers.ones_like(fake_LB_logit))
G_ad_cam_loss_LB = self.MSE_loss(fake_LB_cam_logit,
fluid.layers.ones_like(fake_LB_cam_logit))
#图像重建损失
G_recon_loss_A = self.L1_loss(fake_A2B2A, real_A)
G_recon_loss_B = self.L1_loss(fake_B2A2B, real_B)
#让网络学习到恒等映射，源域和目标域属于同一个域的时候，生成的图片不应该改变它
G_identity_loss_A = self.L1_loss(fake_A2A, real_A)
G_identity_loss_B = self.L1_loss(fake_B2B, real_B)
# tune the feature maps in the decoding part be more like the images from the target domain
# rather than the source domain.
#似乎跟论文公式(5)表达的不是一个意思？公式(5)同域最大化，不同域最小化
G_cam_loss_A = self.BCE_loss(fake_B2A_cam_logit,
fluid.layers.ones_like(fake_B2A_cam_logit))\
+ self.BCE_loss(fake_A2A_cam_logit,
fluid.layers.zeros_like(fake_A2A_cam_logit))
G_cam_loss_B = self.BCE_loss(fake_A2B_cam_logit,fluid.layers.ones_like(fake_A2B_cam_logit)) \
+ self.BCE_loss(fake_B2B_cam_logit,fluid.layers.zeros_like(fake_B2B_cam_logit))
G_loss_A = self.adv_weight * (
G_ad_loss_GA + G_ad_cam_loss_GA + G_ad_loss_LA + G_ad_cam_loss_LA) + self.cycle_weight * G_recon_loss_A + self.identity_weight * G_identity_loss_A + self.cam_weight * G_cam_loss_A
G_loss_B = self.adv_weight * (
G_ad_loss_GB + G_ad_cam_loss_GB + G_ad_loss_LB + G_ad_cam_loss_LB) + self.cycle_weight * G_recon_loss_B + self.identity_weight * G_identity_loss_B + self.cam_weight * G_cam_loss_B
Generator_loss = G_loss_A + G_loss_B
Generator_loss.backward()
self.G_optim.minimize(Generator_loss)
print("[%5d/%5d] time: %4.4f d_loss: %.8f, g_loss: %.8f" % (
step, self.iteration, time.time() - start_time, Discriminator_loss, Generator_loss))
#self.logger.add_scalar(tag="d_loss",step=step,value=Discriminator_loss.numpy())
#self.logger.add_scalar(tag="g_loss", step=step, value=Generator_loss.numpy())
if step % self.print_freq == 0:
train_sample_num = 5
test_sample_num = 5
A2B = np.zeros((self.img_size * 7, 0, 3))
B2A = np.zeros((self.img_size * 7, 0, 3))
self.genA2B.eval(), self.genB2A.eval(), self.disGA.eval(), self.disGB.eval(), self.disLA.eval(), self.disLB.eval()
for _ in range(train_sample_num):
real_A_iter = next(self.trainA_loader())
real_B_iter = next(self.trainB_loader())
real_A, _ = real_A_iter[0][0], real_A_iter[0][1]
real_B, _ = real_B_iter[0][0], real_B_iter[0][1]
real_A = real_A[np.newaxis, :]
real_B = real_B[np.newaxis, :]
real_A, real_B = fluid.dygraph.to_variable(real_A), fluid.dygraph.to_variable(real_B)
fake_A2B, _, fake_A2B_heatmap = self.genA2B(real_A)
fake_B2A, _, fake_B2A_heatmap = self.genB2A(real_B)
fake_A2B2A, _, fake_A2B2A_heatmap = self.genB2A(fake_A2B)
fake_B2A2B, _, fake_B2A2B_heatmap = self.genA2B(fake_B2A)
fake_A2A, _, fake_A2A_heatmap = self.genB2A(real_A)
fake_B2B, _, fake_B2B_heatmap = self.genA2B(real_B)
A2B = np.concatenate((A2B, np.concatenate((RGB2BGR(denorm(real_A[0].numpy().transpose(1,2,0))),
cam(fake_A2A_heatmap[0].numpy().transpose(1,2,0),
self.img_size),
RGB2BGR(denorm(fake_A2A[0].numpy().transpose(1,2,0))),
cam(fake_A2B_heatmap[0].numpy().transpose(1,2,0),
self.img_size),
RGB2BGR(denorm(fake_A2B[0].numpy().transpose(1,2,0))),
cam(fake_A2B2A_heatmap[0].numpy().transpose(1,2,0),
self.img_size),
RGB2BGR(denorm(fake_A2B2A[0].numpy().transpose(1,2,0)))), 0)),
1)
B2A = np.concatenate((B2A, np.concatenate((RGB2BGR(denorm(real_B[0].numpy().transpose(1,2,0))),
cam(fake_B2B_heatmap[0].numpy().transpose(1,2,0),
self.img_size),
RGB2BGR(denorm(fake_B2B[0].numpy().transpose(1,2,0))),
cam(fake_B2A_heatmap[0].numpy().transpose(1,2,0),
self.img_size),
RGB2BGR(denorm(fake_B2A[0].numpy().transpose(1,2,0))),
cam(fake_B2A2B_heatmap[0].numpy().transpose(1,2,0),
self.img_size),
RGB2BGR(denorm(fake_B2A2B[0].numpy().transpose(1,2,0)))), 0)),
1)
for t in range(test_sample_num):
real_A_iter_t = next(self.testA_loader())
real_B_iter_t = next(self.testB_loader())
real_A, _ = real_A_iter_t[0][0], real_A_iter_t[0][1]
real_B, _ = real_B_iter_t[0][0], real_B_iter_t[0][1]
real_A = real_A[np.newaxis, :]
real_B = real_B[np.newaxis, :]
real_A, real_B = fluid.dygraph.to_variable(real_A), fluid.dygraph.to_variable(real_B)
fake_A2B, _, fake_A2B_heatmap = self.genA2B(real_A)
fake_B2A, _, fake_B2A_heatmap = self.genB2A(real_B)
fake_A2B2A, _, fake_A2B2A_heatmap = self.genB2A(fake_A2B)
fake_B2A2B, _, fake_B2A2B_heatmap = self.genA2B(fake_B2A)
fake_A2A, _, fake_A2A_heatmap = self.genB2A(real_A)
fake_B2B, _, fake_B2B_heatmap = self.genA2B(real_B)
A2B = np.concatenate((A2B, np.concatenate((RGB2BGR(denorm(real_A[0].numpy().transpose(1,2,0))),
cam(fake_A2A_heatmap[0].numpy().transpose(1,2,0),
self.img_size),
RGB2BGR(denorm(fake_A2A[0].numpy().transpose(1,2,0))),
cam(fake_A2B_heatmap[0].numpy().transpose(1,2,0),
self.img_size),
RGB2BGR(denorm(fake_A2B[0].numpy().transpose(1,2,0))),
cam(fake_A2B2A_heatmap[0].numpy().transpose(1,2,0),
self.img_size),
RGB2BGR(denorm(fake_A2B2A[0].numpy().transpose(1,2,0)))), 0)),
1)
B2A = np.concatenate((B2A, np.concatenate((RGB2BGR(denorm(real_B[0].numpy().transpose(1,2,0))),
cam(fake_B2B_heatmap[0].numpy().transpose(1,2,0),
self.img_size),
RGB2BGR(denorm(fake_B2B[0].numpy().transpose(1,2,0))),
cam(fake_B2A_heatmap[0].numpy().transpose(1,2,0),
self.img_size),
RGB2BGR(denorm(fake_B2A[0].numpy().transpose(1,2,0))),
cam(fake_B2A2B_heatmap[0].numpy().transpose(1,2,0),
self.img_size),
RGB2BGR(denorm(fake_B2A2B[0].numpy().transpose(1,2,0)))), 0)),
1)
cv2.imwrite(os.path.join(self.result_dir, self.dataset, 'img', 'A2B_%07d.png' % step), A2B * 255.0)
cv2.imwrite(os.path.join(self.result_dir, self.dataset, 'img', 'B2A_%07d.png' % step), B2A * 255.0)
self.genA2B.train(), self.genB2A.train(), self.disGA.train(), self.disGB.train(), self.disLA.train(), self.disLB.train()
if step % self.save_freq == 0:
self.save(os.path.join(self.result_dir, self.dataset), step)
if step % 1000 == 0:
fluid.save_dygraph(self.G_optim.state_dict(),os.path.join(self.result_dir, self.dataset,'model','GOptim_latest'))
fluid.save_dygraph(self.D_optim.state_dict(),os.path.join(self.result_dir, self.dataset, 'model', 'DOptim_latest'))
fluid.save_dygraph(self.genA2B.state_dict(), os.path.join(self.result_dir, self.dataset,'model','genA2B_latest'))
fluid.save_dygraph(self.genB2A.state_dict(), os.path.join(self.result_dir, self.dataset,'model','genB2A_latest'))
fluid.save_dygraph(self.disGA.state_dict(), os.path.join(self.result_dir, self.dataset ,'model','disGA_latest'))
fluid.save_dygraph(self.disGB.state_dict(), os.path.join(self.result_dir, self.dataset ,'model', 'disGB_latest'))
fluid.save_dygraph(self.disLA.state_dict(), os.path.join(self.result_dir, self.dataset ,'model', 'disLA_latest'))
fluid.save_dygraph(self.disLB.state_dict(), os.path.join(self.result_dir, self.dataset ,'model', 'disLB_latest'))
def save(self, res_dir, step):
fluid.save_dygraph(self.G_optim.state_dict(),os.path.join(res_dir, 'GOptim_%07d' % step))
fluid.save_dygraph(self.D_optim.state_dict(),os.path.join(res_dir, 'DOptim_%07d' % step))
fluid.save_dygraph(self.genA2B.state_dict(), os.path.join(res_dir,'genA2B_%07d' % step))
fluid.save_dygraph(self.genB2A.state_dict(), os.path.join(res_dir,'genB2A_%07d' % step))
fluid.save_dygraph(self.disGA.state_dict(), os.path.join(res_dir,'disGA_%07d' % step))
fluid.save_dygraph(self.disGB.state_dict(), os.path.join(res_dir,'disGB_%07d' % step))
fluid.save_dygraph(self.disLA.state_dict(), os.path.join(res_dir,'disLA_%07d' % step))
fluid.save_dygraph(self.disLB.state_dict(), os.path.join(res_dir, 'disLB_%07d' % step))
def load(self, res_dir, step):
genA2B_param,_ = fluid.load_dygraph(os.path.join(res_dir,'genA2B_%07d' % step))
for k,v in genA2B_param.items():
if "FC.2.weight" in k:
genA2B_param["FC.1.weight"] = genA2B_param.pop("FC.2.weight")
break
self.genA2B.set_dict(genA2B_param)
genB2A_param,_ = fluid.load_dygraph(os.path.join(res_dir,'genB2A_%07d' % step))
for k, v in genB2A_param.items():
if "FC.2.weight" in k:
genB2A_param["FC.1.weight"] = genB2A_param.pop("FC.2.weight")
break
self.genB2A.set_dict(genB2A_param)
disGA_param,_ = fluid.load_dygraph(os.path.join(res_dir,'disGA_%07d' % step))
self.disGA.set_dict(disGA_param)
disGB_param,_ = fluid.load_dygraph(os.path.join(res_dir,'disGB_%07d' % step))
self.disGB.set_dict(disGB_param)
disLA_param,_ = fluid.load_dygraph(os.path.join(res_dir,'disLA_%07d' % step))
self.disLA.set_dict(disLA_param)
disLB_param,_ = fluid.load_dygraph(os.path.join(res_dir, 'disLB_%07d' % step))
self.disLB.set_dict(disLB_param)
def load_latest(self, res_dir):
_,gopt_param = fluid.load_dygraph(os.path.join(res_dir,'GOptim_latest'))
self.G_optim.set_dict(gopt_param)
_, dopt_param = fluid.load_dygraph(os.path.join(res_dir, 'DOptim_latest'))
self.D_optim.set_dict(dopt_param)
genA2B_param,_ = fluid.load_dygraph(os.path.join(res_dir,'genA2B_latest'))
for k,v in genA2B_param.items():
if "FC.2.weight" in k:
genA2B_param["FC.1.weight"] = genA2B_param.pop("FC.2.weight")
break
self.genA2B.set_dict(genA2B_param)
genB2A_param,_ = fluid.load_dygraph(os.path.join(res_dir,'genB2A_latest'))
for k,v in genB2A_param.items():
if "FC.2.weight" in k:
genB2A_param["FC.1.weight"] = genB2A_param.pop("FC.2.weight")
break
self.genB2A.set_dict(genB2A_param)
disGA_param,_ = fluid.load_dygraph(os.path.join(res_dir,'disGA_latest'))
self.disGA.set_dict(disGA_param)
disGB_param,_ = fluid.load_dygraph(os.path.join(res_dir,'disGB_latest'))
self.disGB.set_dict(disGB_param)
disLA_param,_ = fluid.load_dygraph(os.path.join(res_dir,'disLA_latest'))
self.disLA.set_dict(disLA_param)
disLB_param,_ = fluid.load_dygraph(os.path.join(res_dir, 'disLB_latest'))
self.disLB.set_dict(disLB_param)
def test(self):
with fluid.dygraph.guard(self.device):
model_list = glob.glob(os.path.join(self.result_dir, self.dataset, '*.*'))
if not len(model_list) == 0:
model_list.sort()
start_iter = int(model_list[-1].split('_')[-1].split('.')[0])
self.load(os.path.join(self.result_dir, self.dataset), start_iter)
print(" [*] Load SUCCESS")
else:
print(" [*] Load FAILURE")
return
self.genA2B.eval(), self.genB2A.eval()
step = 0
if not os.path.exists(os.path.join(self.result_dir, self.dataset, 'fakeA2B')):
os.mkdir(os.path.join(self.result_dir, self.dataset, 'fakeA2B'))
if not os.path.exists(os.path.join(self.result_dir, self.dataset, 'fakeB2A')):
os.mkdir(os.path.join(self.result_dir, self.dataset, 'fakeB2A'))
for data in self.testA_loader():
real_A, _ = data[0][0], data[0][1]
real_A = real_A[np.newaxis, :]
real_A = fluid.dygraph.to_variable(real_A)
fake_A2B, _, _ = self.genA2B(real_A)
fake_A2B = RGB2BGR(denorm(fake_A2B[0].numpy().transpose(1,2,0)))
cv2.imwrite(os.path.join(self.result_dir, self.dataset, 'fakeA2B', 'A2B_%07d.png' % step), fake_A2B*255)
step+=1
step = 0
for data in self.testB_loader():
real_B, _ = data[0][0], data[0][1]
real_B = real_B[np.newaxis, :]
real_B = fluid.dygraph.to_variable(real_B)
fake_B2A, _, _ = self.genB2A(real_B)
fake_B2A = RGB2BGR(denorm(fake_B2A[0].numpy().transpose(1, 2, 0)))
cv2.imwrite(os.path.join(self.result_dir, self.dataset, 'fakeB2A', 'B2A_%07d.png' % step), fake_B2A*255)
step += 1