-
Notifications
You must be signed in to change notification settings - Fork 22
/
Copy pathDCShadowNet_train.py
394 lines (316 loc) · 22.5 KB
/
DCShadowNet_train.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
import time, itertools
from dataset import ImageFolder
from torchvision import transforms
from torch.utils.data import DataLoader
from networks import *
from utils_loss import *
from glob import glob
from PIL import Image
class DCShadowNet(object) :
def __init__(self, args):
self.model_name = 'DCShadowNet'
self.result_dir = args.result_dir
self.dataset = args.dataset
self.datasetpath = args.datasetpath
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.dom_weight = args.dom_weight
self.use_ch_loss = args.use_ch_loss
self.use_pecp_loss = args.use_pecp_loss
self.use_smooth_loss = args.use_smooth_loss
if args.use_ch_loss == True:
self.ch_weight = args.ch_weight
if args.use_pecp_loss == True:
self.pecp_weight = args.pecp_weight
if args.use_smooth_loss == True:
self.smooth_weight = args.smooth_weight
""" Generator """
self.n_res = args.n_res
""" Discriminator """
self.n_dis = args.n_dis
self.img_size = args.img_size
self.img_ch = args.img_ch
self.device = args.device
self.benchmark_flag = args.benchmark_flag
self.resume = args.resume
if torch.backends.cudnn.enabled and self.benchmark_flag:
print('set benchmark !')
torch.backends.cudnn.benchmark = True
print()
print("##### Information #####")
print("# dataset : ", self.dataset)
print("# datasetpath : ", self.datasetpath)
def build_model(self):
""" DataLoader """
train_transform = transforms.Compose([
transforms.RandomHorizontalFlip(),
transforms.Resize((self.img_size + 30, self.img_size+30)),
transforms.RandomCrop(self.img_size),
transforms.ToTensor(),
transforms.Normalize(mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5))
])
test_transform = transforms.Compose([
transforms.Resize((self.img_size, self.img_size)),
transforms.ToTensor(),
transforms.Normalize(mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5))
])
self.trainA = ImageFolder(os.path.join('dataset', self.datasetpath, 'trainA'), train_transform)
self.trainB = ImageFolder(os.path.join('dataset', self.datasetpath, 'trainB'), train_transform)
if self.use_ch_loss:
self.trainC = ImageFolder(os.path.join('dataset', self.datasetpath, 'trainC'), train_transform) ##offline load physics ch_norm
self.testA = ImageFolder(os.path.join('dataset', self.datasetpath, 'testA'), test_transform)
self.testB = ImageFolder(os.path.join('dataset', self.datasetpath, 'testB'), test_transform)
if self.use_ch_loss:
self.testC = ImageFolder(os.path.join('dataset', self.datasetpath, 'testC'), test_transform) ##offline load physics ch_norm
self.trainA_loader = DataLoader(self.trainA, batch_size=self.batch_size, shuffle=True)
self.trainB_loader = DataLoader(self.trainB, batch_size=self.batch_size, shuffle=True)
if self.use_ch_loss:
self.trainC_loader = DataLoader(self.trainC, batch_size=self.batch_size, shuffle=False)
self.testA_loader = DataLoader(self.testA, batch_size=1, shuffle=False)
self.testB_loader = DataLoader(self.testB, batch_size=1, shuffle=False)
if self.use_ch_loss:
self.testC_loader = DataLoader(self.testC, batch_size=1, shuffle=False)
""" 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=True).to(self.device)
self.genB2A = ResnetGenerator(input_nc=3, output_nc=3, ngf=self.ch, n_blocks=self.n_res, img_size=self.img_size, light=True).to(self.device)
self.disGA = Discriminator(input_nc=3, ndf=self.ch, n_layers=7).to(self.device)
self.disGB = Discriminator(input_nc=3, ndf=self.ch, n_layers=7).to(self.device)
self.disLA = Discriminator(input_nc=3, ndf=self.ch, n_layers=5).to(self.device)
self.disLB = Discriminator(input_nc=3, ndf=self.ch, n_layers=5).to(self.device)
""" Define Loss """
self.L1_loss = nn.L1Loss().to(self.device)
self.MSE_loss = nn.MSELoss().to(self.device)
self.BCE_loss = nn.BCEWithLogitsLoss().to(self.device)
""" Trainer """
self.G_optim = torch.optim.Adam(itertools.chain(self.genA2B.parameters(), self.genB2A.parameters()), lr=self.lr, betas=(0.5, 0.999), weight_decay=self.weight_decay)
self.D_optim = torch.optim.Adam(itertools.chain(self.disGA.parameters(), self.disGB.parameters(), self.disLA.parameters(), self.disLB.parameters()), lr=self.lr, betas=(0.5, 0.999), weight_decay=self.weight_decay)
self.Rho_clipper = RhoClipper(0, 1)
def train(self):
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(os.path.join(self.result_dir, self.dataset, 'model', '*.pt'))
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, 'model'), start_iter)
print(" [*] Load SUCCESS")
if self.decay_flag and start_iter > (self.iteration // 2):
self.G_optim.param_groups[0]['lr'] -= (self.lr / (self.iteration // 2)) * (start_iter - self.iteration // 2)
self.D_optim.param_groups[0]['lr'] -= (self.lr / (self.iteration // 2)) * (start_iter - self.iteration // 2)
# training loop
print('training start !')
start_time = time.time()
for step in range(start_iter, self.iteration + 1):
if self.decay_flag and step > (self.iteration // 2):
self.G_optim.param_groups[0]['lr'] -= (self.lr / (self.iteration // 2))
self.D_optim.param_groups[0]['lr'] -= (self.lr / (self.iteration // 2))
try:
real_A, _ = trainA_iter.next()
except:
trainA_iter = iter(self.trainA_loader)
real_A, _ = trainA_iter.next()
try:
real_B, _ = trainB_iter.next()
except:
trainB_iter = iter(self.trainB_loader)
real_B, _ = trainB_iter.next()
if self.use_ch_loss:
try:
real_C, _ = trainC_iter.next()
except:
trainC_iter = iter(self.trainC_loader)
real_C, _ = trainC_iter.next()
real_A, real_B = real_A.to(self.device), real_B.to(self.device)
if self.use_ch_loss:
real_C = real_C.to(self.device)
# Update D
self.D_optim.zero_grad()
fake_A2B, _, _ = self.genA2B(real_A)
fake_B2A, _, _ = self.genB2A(real_B)
real_GA_logit, real_GA_Dom_logit, _ = self.disGA(real_A)
real_LA_logit, real_LA_Dom_logit, _ = self.disLA(real_A)
real_GB_logit, real_GB_Dom_logit, _ = self.disGB(real_B)
real_LB_logit, real_LB_Dom_logit, _ = self.disLB(real_B)
fake_GA_logit, fake_GA_Dom_logit, _ = self.disGA(fake_B2A)
fake_LA_logit, fake_LA_Dom_logit, _ = self.disLA(fake_B2A)
fake_GB_logit, fake_GB_Dom_logit, _ = self.disGB(fake_A2B)
fake_LB_logit, fake_LB_Dom_logit, _ = self.disLB(fake_A2B)
D_ad_loss_GA = self.MSE_loss(real_GA_logit, torch.ones_like(real_GA_logit).to(self.device)) + self.MSE_loss(fake_GA_logit, torch.zeros_like(fake_GA_logit).to(self.device))
D_ad_Dom_loss_GA = self.MSE_loss(real_GA_Dom_logit, torch.ones_like(real_GA_Dom_logit).to(self.device)) + self.MSE_loss(fake_GA_Dom_logit, torch.zeros_like(fake_GA_Dom_logit).to(self.device))
D_ad_loss_LA = self.MSE_loss(real_LA_logit, torch.ones_like(real_LA_logit).to(self.device)) + self.MSE_loss(fake_LA_logit, torch.zeros_like(fake_LA_logit).to(self.device))
D_ad_Dom_loss_LA = self.MSE_loss(real_LA_Dom_logit, torch.ones_like(real_LA_Dom_logit).to(self.device)) + self.MSE_loss(fake_LA_Dom_logit, torch.zeros_like(fake_LA_Dom_logit).to(self.device))
D_ad_loss_GB = self.MSE_loss(real_GB_logit, torch.ones_like(real_GB_logit).to(self.device)) + self.MSE_loss(fake_GB_logit, torch.zeros_like(fake_GB_logit).to(self.device))
D_ad_Dom_loss_GB = self.MSE_loss(real_GB_Dom_logit, torch.ones_like(real_GB_Dom_logit).to(self.device)) + self.MSE_loss(fake_GB_Dom_logit, torch.zeros_like(fake_GB_Dom_logit).to(self.device))
D_ad_loss_LB = self.MSE_loss(real_LB_logit, torch.ones_like(real_LB_logit).to(self.device)) + self.MSE_loss(fake_LB_logit, torch.zeros_like(fake_LB_logit).to(self.device))
D_ad_Dom_loss_LB = self.MSE_loss(real_LB_Dom_logit, torch.ones_like(real_LB_Dom_logit).to(self.device)) + self.MSE_loss(fake_LB_Dom_logit, torch.zeros_like(fake_LB_Dom_logit).to(self.device))
D_loss_A = self.adv_weight * (D_ad_loss_GA + D_ad_Dom_loss_GA + D_ad_loss_LA + D_ad_Dom_loss_LA)
D_loss_B = self.adv_weight * (D_ad_loss_GB + D_ad_Dom_loss_GB + D_ad_loss_LB + D_ad_Dom_loss_LB)
Discriminator_loss = D_loss_A + D_loss_B
Discriminator_loss.backward()
self.D_optim.step()
# Update G
self.G_optim.zero_grad()
fake_A2B, fake_A2B_Dom_logit, _ = self.genA2B(real_A) ##yy: fake_B = netG_A2B(real_A), Gb(a)
fake_B2A, fake_B2A_Dom_logit, _ = self.genB2A(real_B) ##yy: fake_A = netG_B2A(real_B), Ga(b)
fake_A2B2A, _, _ = self.genB2A(fake_A2B) ##yy: recovered_A = netG_B2A(fake_B), Ga(Gb(a))
fake_B2A2B, _, _ = self.genA2B(fake_B2A) ##yy: recovered_B = netG_A2B(fake_A), Gb(Ga(b))
fake_A2A, fake_A2A_Dom_logit, _ = self.genB2A(real_A) #yy: G_B2A(A) should equal A if real_A is fed, same_A = netG_B2A(real_A), same_A is fake_A2A
fake_B2B, fake_B2B_Dom_logit, _ = self.genA2B(real_B) #yy: G_A2B(B) should equal B if real_B is fed, same_B = netG_A2B(real_B), same_B is fake_B2B
fake_GA_logit, fake_GA_Dom_logit, _ = self.disGA(fake_B2A) #Da(Ga(b))_global
fake_LA_logit, fake_LA_Dom_logit, _ = self.disLA(fake_B2A) #Da(Ga(b))_local
fake_GB_logit, fake_GB_Dom_logit, _ = self.disGB(fake_A2B) #Db(Gb(a))_global
fake_LB_logit, fake_LB_Dom_logit, _ = self.disLB(fake_A2B) #Db(Gb(a))_local
G_ad_loss_GA = self.MSE_loss(fake_GA_logit, torch.ones_like(fake_GA_logit).to(self.device)) ##yy: log(Da(Ga(b))), loss_GAN_B2A = criterion_GAN(pred_fake, target_real), global D
G_ad_Dom_loss_GA = self.MSE_loss(fake_GA_Dom_logit, torch.ones_like(fake_GA_Dom_logit).to(self.device)) ##yy: G Dom
G_ad_loss_LA = self.MSE_loss(fake_LA_logit, torch.ones_like(fake_LA_logit).to(self.device)) ##yy: log(Da(Ga(b))), local D
G_ad_Dom_loss_LA = self.MSE_loss(fake_LA_Dom_logit, torch.ones_like(fake_LA_Dom_logit).to(self.device)) ##yy: L Dom
G_ad_loss_GB = self.MSE_loss(fake_GB_logit, torch.ones_like(fake_GB_logit).to(self.device)) ##yy: log(Db(Gb(a))), loss_GAN_A2B = criterion_GAN(pred_fake, target_real), global D
G_ad_Dom_loss_GB = self.MSE_loss(fake_GB_Dom_logit, torch.ones_like(fake_GB_Dom_logit).to(self.device))
G_ad_loss_LB = self.MSE_loss(fake_LB_logit, torch.ones_like(fake_LB_logit).to(self.device)) ##yy: log(Db(Gb(a))), local D
G_ad_Dom_loss_LB = self.MSE_loss(fake_LB_Dom_logit, torch.ones_like(fake_LB_Dom_logit).to(self.device))
G_recon_loss_A = self.L1_loss(fake_A2B2A, real_A) #yy: ||Ga(Gb(a))-a||1, loss_cycle_ABA
G_recon_loss_B = self.L1_loss(fake_B2A2B, real_B) #yy: ||Gb(Ga(b))-b||1, loss_cycle_BAB
G_identity_loss_A = self.L1_loss(fake_A2A, real_A) #yy: ||Ga(a)-a||1, loss_identity_A
G_identity_loss_B = self.L1_loss(fake_B2B, real_B) #yy: ||Gb(b)-b||1, loss_identity_B
##Binary Domain-Classifier
G_dom_loss_A = self.BCE_loss(fake_B2A_Dom_logit, torch.ones_like(fake_B2A_Dom_logit).to(self.device)) + self.BCE_loss(fake_A2A_Dom_logit, torch.zeros_like(fake_A2A_Dom_logit).to(self.device)) ##fake_A, 1; same_A(fake_A2A) 0
G_dom_loss_B = self.BCE_loss(fake_A2B_Dom_logit, torch.ones_like(fake_A2B_Dom_logit).to(self.device)) + self.BCE_loss(fake_B2B_Dom_logit, torch.zeros_like(fake_B2B_Dom_logit).to(self.device)) ##fake_B, 1; same_B(fake_B2B) 0
if self.use_pecp_loss:
selfpecpvgg_loss = PerceptualLossVgg16(None,
[0],
weights=[1.0],
indices=[22]) ##yy: Checkpoint/Model for pecp_vgg loss, ImageNet, layer 22
loss_selfpecp = selfpecpvgg_loss(fake_A2B, real_A)
if self.use_smooth_loss:
gen_mask = softmask_generator(real_A, fake_A2B)
loss_smooth = smooth_loss_masked(fake_A2B, gen_mask)
if self.use_ch_loss:
fake_A2B_ = (fake_A2B+1.)/2.
ch_z = fake_A2B_/ fake_A2B_.sum(dim=1, keepdim=True).clamp(min=1e-8) ##yy: for fake_A2B do chromaticity operation c/(r+g+b) to get ch_z
ch_z = 2*ch_z-1
ch_norm = real_C ##yy: offline load ch_norm
loss_ch = self.L1_loss(ch_z, ch_norm)
G_loss_A = self.adv_weight * (G_ad_loss_GA + G_ad_Dom_loss_GA + G_ad_loss_LA + G_ad_Dom_loss_LA) + self.cycle_weight * G_recon_loss_A + self.identity_weight * G_identity_loss_A + self.dom_weight * G_dom_loss_A
G_loss_B = self.adv_weight * (G_ad_loss_GB + G_ad_Dom_loss_GB + G_ad_loss_LB + G_ad_Dom_loss_LB) + self.cycle_weight * G_recon_loss_B + self.identity_weight * G_identity_loss_B + self.dom_weight * G_dom_loss_B
Generator_loss = G_loss_A + G_loss_B
if self.use_ch_loss == True:
Generator_loss = Generator_loss + loss_ch
if self.use_pecp_loss == True:
Generator_loss = Generator_loss + loss_selfpecp
if self.use_smooth_loss == True:
Generator_loss = Generator_loss + loss_smooth
Generator_loss.backward()
self.G_optim.step()
self.genA2B.apply(self.Rho_clipper)
self.genB2A.apply(self.Rho_clipper)
print("[%5d/%5d] time: %4.4f d_loss: %.8f, g_loss: %.8f" % (step, self.iteration, time.time() - start_time, Discriminator_loss, Generator_loss))
with torch.no_grad():
if step % self.print_freq == 0:
train_sample_num = 5
test_sample_num = 5
A2B = np.zeros((self.img_size * 4, 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):
try:
real_A, _ = trainA_iter.next()
except:
trainA_iter = iter(self.trainA_loader)
real_A, _ = trainA_iter.next()
try:
real_B, _ = trainB_iter.next()
except:
trainB_iter = iter(self.trainB_loader)
real_B, _ = trainB_iter.next()
real_A, real_B = real_A.to(self.device), real_B.to(self.device)
fake_A2B, _, _ = self.genA2B(real_A)
fake_B2A, _, _ = self.genB2A(real_B)
fake_A2B2A, _, _ = self.genB2A(fake_A2B) ##yy: recovered_A = netG_B2A(fake_B), Ga(Gb(a))
fake_B2A2B, _, _ = self.genA2B(fake_B2A) ##yy: recovered_B = netG_A2B(fake_A), Gb(Ga(b))
fake_A2A, _, _ = self.genB2A(real_A) #yy: G_B2A(A) should equal A if real A is fed, same_A = netG_B2A(real_A), same_A(fake_A2A)
fake_B2B, _, _ = self.genA2B(real_B) #yy: G_A2B(B) should equal B if real B is fed, same_B = netG_A2B(real_B), same_B(fake_B2B)
A2B = np.concatenate((A2B, np.concatenate((RGB2BGR(tensor2numpy(denorm(real_A[0]))),
RGB2BGR(tensor2numpy(denorm(fake_A2B[0]))),
RGB2BGR(tensor2numpy(denorm(fake_A2A[0]))),
RGB2BGR(tensor2numpy(denorm(fake_A2B2A[0])))), 0)), 1)
for _ in range(test_sample_num):
try:
real_A, _ = testA_iter.next()
except:
testA_iter = iter(self.testA_loader)
real_A, _ = testA_iter.next()
try:
real_B, _ = testB_iter.next()
except:
testB_iter = iter(self.testB_loader)
real_B, _ = testB_iter.next()
if self.use_ch_loss:
try:
real_C_test, _ = testC_iter.next()
except:
testC_iter = iter(self.testC_loader)
real_C_test, _ = testC_iter.next()
real_A, real_B = real_A.to(self.device), real_B.to(self.device)
if self.use_ch_loss:
real_C_test = real_C_test.to(self.device)
fake_A2B, _, _ = self.genA2B(real_A)
fake_B2A, _, _ = self.genB2A(real_B)
fake_A2B2A, _, _ = self.genB2A(fake_A2B)
fake_B2A2B, _, _ = self.genA2B(fake_B2A)
fake_A2A, _, _ = self.genB2A(real_A)
fake_B2B, _, _ = self.genA2B(real_B)
if self.use_smooth_loss == True:
gen_mask = softmask_generator(real_A, fake_A2B)
A2B = np.concatenate((A2B, np.concatenate((RGB2BGR(tensor2numpy(denorm(real_A[0]))),
RGB2BGR(tensor2numpy(denorm(fake_A2B[0]))),
RGB2BGR(tensor2numpy(denorm(gen_mask[0]))),
RGB2BGR(tensor2numpy(denorm(fake_A2B2A[0])))), 0)), 1)
if self.use_ch_loss == True:
fake_A2B_ = (fake_A2B+1.)/2.
ch_z = fake_A2B_/ fake_A2B_.sum(dim=1, keepdim=True).clamp(min=1e-8)
ch_z_test = 2*ch_z-1
ch_norm_test = real_C_test
A2B = np.concatenate((A2B, np.concatenate((RGB2BGR(tensor2numpy(denorm(real_A[0]))),
RGB2BGR(tensor2numpy(denorm(fake_A2B[0]))),
RGB2BGR(tensor2numpy(denorm(ch_norm_test[0]))),
RGB2BGR(tensor2numpy(denorm(ch_z_test[0])))), 0)), 1)
cv2.imwrite(os.path.join(self.result_dir, self.dataset, 'train_img', 'A2B_%07d.png' % step), A2B * 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, 'model'), step)
if step % 1000 == 0:
params = {}
params['genA2B'] = self.genA2B.state_dict()
params['genB2A'] = self.genB2A.state_dict()
params['disGA'] = self.disGA.state_dict()
params['disGB'] = self.disGB.state_dict()
params['disLA'] = self.disLA.state_dict()
params['disLB'] = self.disLB.state_dict()
torch.save(params, os.path.join(self.result_dir, self.dataset + '_params_latest.pt'))
def save(self, dir, step):
params = {}
params['genA2B'] = self.genA2B.state_dict()
params['genB2A'] = self.genB2A.state_dict()
params['disGA'] = self.disGA.state_dict()
params['disGB'] = self.disGB.state_dict()
params['disLA'] = self.disLA.state_dict()
params['disLB'] = self.disLB.state_dict()
torch.save(params, os.path.join(dir, self.dataset + '_params_%07d.pt' % step))
def load(self, dir, step):
params = torch.load(os.path.join(dir, self.dataset + '_params_%07d.pt' % step))
self.genA2B.load_state_dict(params['genA2B'])
self.genB2A.load_state_dict(params['genB2A'])
self.disGA.load_state_dict(params['disGA'])
self.disGB.load_state_dict(params['disGB'])
self.disLA.load_state_dict(params['disLA'])
self.disLB.load_state_dict(params['disLB'])