-
Notifications
You must be signed in to change notification settings - Fork 2
/
cross_teacher_contr.py
1321 lines (1017 loc) · 52.5 KB
/
cross_teacher_contr.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
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
import argparse
import os
import sys
import random
import timeit
import datetime
import numpy as np
import pickle
import scipy.misc
import torch
import torch.nn as nn
import torch.optim as optim
import torch.nn.functional as F
import torch.backends.cudnn as cudnn
from torch.utils import data, model_zoo
from torch.autograd import Variable
import torchvision.transforms as transform
from model.deeplabv2 import Deeplab as Res_Deeplab
from utils.loss import CrossEntropy2d
from utils.loss import CrossEntropyLoss2dPixelWiseWeighted
from utils.loss import MSELoss2d
from utils.loss import CtsEntropy2d
from utils import transformmasks
from utils import transformsgpu
from utils.helpers import colorize_mask
import utils.palette as palette
from utils.sync_batchnorm import convert_model
from utils.sync_batchnorm import DataParallelWithCallback
from data.voc_dataset import VOCDataSet
from data import get_loader, get_data_path
from data.augmentations import *
from tqdm import tqdm
import PIL
from torchvision import transforms
import json
from torch.utils import tensorboard
from evaluateSSL import evaluate
from utils.method import get_accuracy, get_meaniou
#from apex import amp
import time
start = timeit.default_timer()
start_writeable = datetime.datetime.now().strftime('%m-%d_%H-%M')
def get_arguments():
"""Parse all the arguments provided from the CLI.
Returns:
A list of parsed arguments.
"""
parser = argparse.ArgumentParser(description="DeepLab-ResNet Network")
parser.add_argument("--gpus", type=int, default=1,
help="choose number of gpu devices to use (default: 1)")
parser.add_argument("-c", "--config", type=str, default='config.json',
help='Path to the config file (default: config.json)')
parser.add_argument("-r", "--resume", type=str, default=None,
help='Path to the .pth file to resume from (default: None)')
parser.add_argument("-n", "--name", type=str, default=None, required=True,
help='Name of the run (default: None)')
parser.add_argument("--save-images", type=str, default=None,
help='Include to save images (default: None)')
return parser.parse_args()
def loss_calc(pred, label):
label = Variable(label.long()).cuda()
if len(gpus) > 1:
criterion = torch.nn.DataParallel(CrossEntropy2d(ignore_label=ignore_label),
device_ids=gpus).cuda() # Ignore label ??
else:
criterion = CrossEntropy2d(ignore_label=ignore_label).cuda() # Ignore label ??
return criterion(pred, label)
def loss_cts(pred, label):
label = Variable(label.long()).cuda()
if len(gpus) > 1:
criterion = torch.nn.DataParallel(CtsEntropy2d(ignore_label=ignore_label),
device_ids=gpus).cuda() # Ignore label ??
else:
criterion = CtsEntropy2d(ignore_label=ignore_label).cuda() # Ignore label ??
return criterion(pred, label)
def ent_loss(pred):
pred_logsoftmax = torch.log_softmax(pred, dim=1)
pred_softmax = torch.softmax(pred, dim=1)
loss = torch.sum(-pred_softmax * pred_logsoftmax) / (pred.size(0) * pred.size(2) * pred.size(3))
return loss
def ema_loss(model, ema_model, images_remain, inputs_u_w, weak_parameters, interp, i_iter, unlabeled_loss):
with torch.no_grad():
logits_u_w = interp(ema_model(inputs_u_w.detach())['out'])
logits_u_w, _ = weakTransform(getWeakInverseTransformParameters(weak_parameters), data=logits_u_w.detach())
softmax_u_w = torch.softmax(logits_u_w.detach(), dim=1)
max_probs, argmax_u_w = torch.max(softmax_u_w, dim=1)
if mix_mask == "class":
for image_i in range(batch_size):
classes = torch.unique(argmax_u_w[image_i])
classes = classes[classes != ignore_label]
nclasses = classes.shape[0]
classes = (classes[torch.Tensor(
np.random.choice(nclasses, int((nclasses - nclasses % 2) / 2), replace=False)).long()]).cuda()
if image_i == 0:
MixMask = transformmasks.generate_class_mask(argmax_u_w[image_i], classes).unsqueeze(0).cuda()
else:
MixMask = torch.cat(
(MixMask, transformmasks.generate_class_mask(argmax_u_w[image_i], classes).unsqueeze(0).cuda()))
elif mix_mask == 'cut':
img_size = inputs_u_w.shape[2:4]
for image_i in range(batch_size):
if image_i == 0:
MixMask = torch.from_numpy(transformmasks.generate_cutout_mask(img_size)).unsqueeze(0).cuda().float()
else:
MixMask = torch.cat((MixMask, torch.from_numpy(transformmasks.generate_cutout_mask(img_size)).unsqueeze(
0).cuda().float()))
elif mix_mask == "cow":
img_size = inputs_u_w.shape[2:4]
sigma_min = 8
sigma_max = 32
p_min = 0.5
p_max = 0.5
for image_i in range(batch_size):
sigma = np.exp(np.random.uniform(np.log(sigma_min), np.log(sigma_max))) # Random sigma
p = np.random.uniform(p_min, p_max) # Random p
if image_i == 0:
MixMask = torch.from_numpy(transformmasks.generate_cow_mask(img_size, sigma, p, seed=None)).unsqueeze(
0).cuda().float()
else:
MixMask = torch.cat((MixMask, torch.from_numpy(
transformmasks.generate_cow_mask(img_size, sigma, p, seed=None)).unsqueeze(0).cuda().float()))
elif mix_mask == None:
MixMask = torch.ones((inputs_u_w.shape)).cuda()
strong_parameters = {"Mix": MixMask}
#strong_parameters = {}
if random_flip:
strong_parameters["flip"] = random.randint(0, 1)
else:
strong_parameters["flip"] = 0
if color_jitter:
strong_parameters["ColorJitter"] = random.uniform(0, 1)
else:
strong_parameters["ColorJitter"] = 0
if gaussian_blur:
strong_parameters["GaussianBlur"] = random.uniform(0, 1)
else:
strong_parameters["GaussianBlur"] = 0
inputs_u_s, _ = strongTransform(strong_parameters, data=images_remain)
logits_u_p=model(inputs_u_s)
logits_u_s = interp(logits_u_p['out'])
softmax_u_w_mixed, _ = strongTransform(strong_parameters, data=softmax_u_w)
max_probs, pseudo_label = torch.max(softmax_u_w_mixed, dim=1)
if pixel_weight == "threshold_uniform":
unlabeled_weight = torch.sum(max_probs.ge(0.968).long() == 1).item() / np.size(np.array(pseudo_label.cpu()))
pixelWiseWeight = unlabeled_weight * torch.ones(max_probs.shape).cuda()
elif pixel_weight == "threshold":
# pixelWiseWeight = max_probs.ge(0.968).long().cuda()
pixelWiseWeight=torch.ones(max_probs.shape).cuda()
pixelWiseWeight=pixelWiseWeight*torch.pow(max_probs.detach(),6)
elif pixel_weight == 'sigmoid':
max_iter = 10000
pixelWiseWeight = sigmoid_ramp_up(i_iter, max_iter) * torch.ones(max_probs.shape).cuda()
elif pixel_weight == False:
pixelWiseWeight = torch.ones(max_probs.shape).cuda()
if consistency_loss == 'CE':
#labels = torch.cat([pseudo_label.unsqueeze(1).float(), max_probs.unsqueeze(1).float()], dim=1)
L_u = consistency_weight * unlabeled_loss(logits_u_s, pseudo_label, pixelWiseWeight)
#L_u = consistency_weight * unlabeled_loss(logits_u_s, labels,max_probs.shape[0])
elif consistency_loss == 'MSE':
unlabeled_weight = torch.sum(max_probs.ge(0.968).long() == 1).item() / np.size(np.array(pseudo_label.cpu()))
# softmax_u_w_mixed = torch.cat((softmax_u_w_mixed[1].unsqueeze(0),softmax_u_w_mixed[0].unsqueeze(0)))
L_u = consistency_weight * unlabeled_weight * unlabeled_loss(logits_u_s, softmax_u_w_mixed)
del logits_u_w,max_probs,argmax_u_w,MixMask,softmax_u_w_mixed,softmax_u_w,strong_parameters
prob = F.interpolate(max_probs.clone().unsqueeze(1), size=logits_u_p['feat'].shape[2:],mode='nearest').squeeze(1)
max = F.interpolate(pseudo_label.clone().unsqueeze(1).float(), size=logits_u_p['feat'].shape[2:], mode='nearest').squeeze(1).long()
return L_u,logits_u_p,prob,max
def get_in(ema_model,inputs_u_w,images_remain,weak_parameters,interp):
with torch.no_grad():
logits_u_w = interp(ema_model(inputs_u_w.detach())['out'])
logits_u_w, _ = weakTransform(getWeakInverseTransformParameters(weak_parameters), data=logits_u_w.detach())
softmax_u_w = torch.softmax(logits_u_w.detach(), dim=1)
max_probs, argmax_u_w = torch.max(softmax_u_w, dim=1)
if mix_mask == "class":
for image_i in range(batch_size):
classes = torch.unique(argmax_u_w[image_i])
classes = classes[classes != ignore_label]
nclasses = classes.shape[0]
classes = (classes[torch.Tensor(
np.random.choice(nclasses, int((nclasses - nclasses % 2) / 2), replace=False)).long()]).cuda()
if image_i == 0:
MixMask = transformmasks.generate_class_mask(argmax_u_w[image_i], classes).unsqueeze(0).cuda()
else:
MixMask = torch.cat(
(MixMask, transformmasks.generate_class_mask(argmax_u_w[image_i], classes).unsqueeze(0).cuda()))
elif mix_mask == 'cut':
img_size = inputs_u_w.shape[2:4]
for image_i in range(batch_size):
if image_i == 0:
MixMask = torch.from_numpy(transformmasks.generate_cutout_mask(img_size)).unsqueeze(0).cuda().float()
else:
MixMask = torch.cat((MixMask, torch.from_numpy(transformmasks.generate_cutout_mask(img_size)).unsqueeze(
0).cuda().float()))
elif mix_mask == "cow":
img_size = inputs_u_w.shape[2:4]
sigma_min = 8
sigma_max = 32
p_min = 0.5
p_max = 0.5
for image_i in range(batch_size):
sigma = np.exp(np.random.uniform(np.log(sigma_min), np.log(sigma_max))) # Random sigma
p = np.random.uniform(p_min, p_max) # Random p
if image_i == 0:
MixMask = torch.from_numpy(transformmasks.generate_cow_mask(img_size, sigma, p, seed=None)).unsqueeze(
0).cuda().float()
else:
MixMask = torch.cat((MixMask, torch.from_numpy(
transformmasks.generate_cow_mask(img_size, sigma, p, seed=None)).unsqueeze(0).cuda().float()))
elif mix_mask == None:
MixMask = torch.ones((inputs_u_w.shape)).cuda()
#strong_parameters = {"Mix": MixMask}
strong_parameters = {}
if random_flip:
strong_parameters["flip"] = random.randint(0, 1)
else:
strong_parameters["flip"] = 0
if color_jitter:
strong_parameters["ColorJitter"] = random.uniform(0, 1)
else:
strong_parameters["ColorJitter"] = 0
if gaussian_blur:
strong_parameters["GaussianBlur"] = random.uniform(0, 1)
else:
strong_parameters["GaussianBlur"] = 0
inputs_u_s, _ = strongTransform(strong_parameters, data=images_remain)
softmax_u_w_mixed, _ = strongTransform(strong_parameters, data=softmax_u_w)
max_probs, pseudo_label = torch.max(softmax_u_w_mixed, dim=1)
return inputs_u_s,max_probs,pseudo_label
def dual_teacher(model_l,model_r, ema_model_l,ema_model_r, images_remain, inputs_u_w, weak_parameters, interp, i_iter, unlabeled_loss):
inputs_u_s_l, max_probs_l, pseudo_label_l=get_in(ema_model_l,inputs_u_w,images_remain,weak_parameters,interp)
inputs_u_s_r, max_probs_r, pseudo_label_r = get_in(ema_model_r, inputs_u_w, images_remain, weak_parameters, interp)
logits_u_s_l=interp(model_l(inputs_u_s_l)['out'])
logits_u_s_r = interp(model_r(inputs_u_s_r)['out'])
pixelWiseWeight = torch.ones(max_probs_l.shape).cuda()
#labels = torch.cat([pseudo_label.unsqueeze(1).float(), max_probs.unsqueeze(1).float()], dim=1)
L_u_l = consistency_weight * unlabeled_loss(logits_u_s_l, pseudo_label_l, pixelWiseWeight)+consistency_weight * unlabeled_loss(logits_u_s_l, pseudo_label_r, pixelWiseWeight)
L_u_r = consistency_weight * unlabeled_loss(logits_u_s_r, pseudo_label_r, pixelWiseWeight)+consistency_weight * unlabeled_loss(logits_u_s_r, pseudo_label_l, pixelWiseWeight)
#L_u = consistency_weight * unlabeled_loss(logits_u_s, labels,max_probs.shape[0])
return L_u_l,L_u_r
def lr_poly(base_lr, iter, max_iter, power):
return base_lr * ((1 - float(iter) / max_iter) ** (power))
def adjust_learning_rate(optimizer, i_iter):
lr = lr_poly(learning_rate, i_iter, num_iterations, lr_power)
optimizer.param_groups[0]['lr'] = lr
if len(optimizer.param_groups) > 1:
optimizer.param_groups[1]['lr'] = lr * 10
def sigmoid_ramp_up(iter, max_iter):
if iter >= max_iter:
return 1
else:
return np.exp(- 5 * (1 - iter / max_iter) ** 2)
def create_ema_model(model):
ema_model = Res_Deeplab(num_classes=num_classes)
for param in ema_model.parameters():
param.detach_()
mp = list(model.parameters())
mcp = list(ema_model.parameters())
n = len(mp)
for i in range(0, n):
mcp[i].data[:] = mp[i].data[:].clone()
if len(gpus) > 1:
if use_sync_batchnorm:
ema_model = convert_model(ema_model)
ema_model = DataParallelWithCallback(ema_model, device_ids=gpus)
else:
ema_model = torch.nn.DataParallel(ema_model, device_ids=gpus)
return ema_model
def update_ema_variables(ema_model, model, alpha_teacher, iteration):
# Use the "true" average until the exponential average is more correct
alpha_teacher = min(1 - 1 / (iteration + 1), alpha_teacher)
if len(gpus) > 1:
for ema_param, param in zip(ema_model.module.parameters(), model.module.parameters()):
ema_param.data[:] = alpha_teacher * ema_param[:].data[:] + (1 - alpha_teacher) * param[:].data[:]
else:
for ema_param, param in zip(ema_model.parameters(), model.parameters()):
ema_param.data[:] = alpha_teacher * ema_param[:].data[:] + (1 - alpha_teacher) * param[:].data[:]
return ema_model
def strongTransform(parameters, data=None, target=None):
assert ((data is not None) or (target is not None))
if 'Mix' in parameters.keys():
data, target = transformsgpu.mix(mask=parameters["Mix"], data=data, target=target)
data, target = transformsgpu.colorJitter(colorJitter=parameters["ColorJitter"],
img_mean=torch.from_numpy(IMG_MEAN.copy()).cuda(), data=data,
target=target)
data, target = transformsgpu.gaussian_blur(blur=parameters["GaussianBlur"], data=data, target=None)
data, target = transformsgpu.flip(flip=parameters["flip"], data=data, target=target)
return data, target
def weakTransform(parameters, data=None, target=None):
data, target = transformsgpu.flip(flip=parameters["flip"], data=data, target=target)
return data, target
def getWeakInverseTransformParameters(parameters):
return parameters
def getStrongInverseTransformParameters(parameters):
return parameters
class DeNormalize(object):
def __init__(self, mean):
self.mean = mean
def __call__(self, tensor):
IMG_MEAN = torch.from_numpy(self.mean.copy())
IMG_MEAN, _ = torch.broadcast_tensors(IMG_MEAN.unsqueeze(1).unsqueeze(2), tensor)
tensor = tensor + IMG_MEAN
tensor = (tensor / 255).float()
tensor = torch.flip(tensor, (0,))
return tensor
class Learning_Rate_Object(object):
def __init__(self, learning_rate):
self.learning_rate = learning_rate
def save_image(image, epoch, id, palette):
with torch.no_grad():
if image.shape[0] == 3:
restore_transform = transforms.Compose([
DeNormalize(IMG_MEAN),
transforms.ToPILImage()])
image = restore_transform(image)
# image = PIL.Image.fromarray(np.array(image)[:, :, ::-1]) # BGR->RGB
image.save(os.path.join('../visualiseImages/', str(epoch) + id + '.png'))
else:
mask = image.numpy()
colorized_mask = colorize_mask(mask, palette)
colorized_mask.save(os.path.join('../visualiseImages/', str(epoch) + id + '.png'))
def _save_checkpoint(iteration, model, optimizer, config, ema_model, model_name, save_best=False, overwrite=True):
checkpoint = {
'iteration': iteration,
'optimizer': optimizer.state_dict(),
'config': config,
}
if len(gpus) > 1:
checkpoint['model'] = model.module.state_dict()
if train_unlabeled:
checkpoint['ema_model'] = ema_model.module.state_dict()
else:
checkpoint['model'] = model.state_dict()
if train_unlabeled:
checkpoint['ema_model'] = ema_model.state_dict()
if save_best:
filename = os.path.join(checkpoint_dir, f'best_model.pth')
torch.save(checkpoint, filename)
print("Saving current best model: best_model.pth")
else:
filename = os.path.join(checkpoint_dir, f'checkpoint-iter{iteration}.pth')
print(f'\nSaving a checkpoint: {filename} ...')
torch.save(checkpoint, filename)
if overwrite:
try:
os.remove(os.path.join(checkpoint_dir,
f'checkpoint-iter{iteration - save_checkpoint_every - model_name}.pth'))
except:
pass
def _resume_checkpoint(resume_path, model, optimizer, ema_model):
print(f'Loading checkpoint : {resume_path}')
checkpoint = torch.load(resume_path)
# Load last run info, the model params, the optimizer and the loggers
iteration = checkpoint['iteration'] + 1
print('Starting at iteration: ' + str(iteration))
if len(gpus) > 1:
model.module.load_state_dict(checkpoint['model'])
else:
model.load_state_dict(checkpoint['model'])
optimizer.load_state_dict(checkpoint['optimizer'])
if train_unlabeled:
if len(gpus) > 1:
ema_model.module.load_state_dict(checkpoint['ema_model'])
else:
ema_model.load_state_dict(checkpoint['ema_model'])
return iteration, model, optimizer, ema_model
def generate_class_pseudo_labels_every_iter(pseudo_probability, pseudo_label, is_print, class_sampling):
probabilities = [np.array([], dtype=np.float32) for _ in range(num_classes)]
for j in range(num_classes):
probabilities[j] = np.concatenate((probabilities[j], pseudo_probability[pseudo_label == j].cpu().numpy()))
kc = []
for j in range(num_classes):
probabilities[j].sort()
probabilities[j] = probabilities[j][::-1]
if len(probabilities[j]) == 0:
kc.append(0.00001)
else:
kc.append(probabilities[j][int(len(probabilities[j]) * class_sampling[j]) - 1])
del probabilities # Better be safe than...
if is_print == 1:
print(kc)
return kc
every_memory_count = 256
min_add_count = 1
#memory bank update
def add_to_memory_bank(memory_bank,memory_bank_size,pred_rep,labels,num_classes):
pred=pred_rep['out']
rep=pred_rep['feat']
prob,max=torch.max(pred,dim=1)
mask_acc=(labels==max)&(prob>=0.95)
#mask_acc=(labels==max)
labels_unique=torch.unique(labels)
for i in range(num_classes):
if i in labels_unique:
mask=(labels==i)&mask_acc
rep_i=rep.permute(0,2,3,1)[mask]
if int((every_memory_count/labeled_samples)*2)>min_add_count*batch_size:
add_new_count=int((every_memory_count/labeled_samples)*2)
else:
add_new_count=min_add_count*batch_size
#使用top k算法
if len(rep_i)>=add_new_count:
max_sort=prob[mask].sort(descending=True)[1]
rep_i=rep_i[max_sort][:add_new_count]
else:
add_new_count=len(rep_i)
if memory_bank_size[i]+add_new_count<=every_memory_count:
memory_bank[i][memory_bank_size[i]:memory_bank_size[i]+add_new_count]=rep_i
memory_bank_size[i]+=add_new_count
else:
memory_bank_i=memory_bank[i][add_new_count-(every_memory_count-memory_bank_size[i]):memory_bank_size[i]].clone()
memory_bank[i][:every_memory_count-add_new_count]=memory_bank_i
memory_bank[i][every_memory_count-add_new_count:]=rep_i
memory_bank_size[i]=every_memory_count
return memory_bank,memory_bank_size
#contr loss
def contr_loss(memory_bank, pred_rep, labels, mask, which_memory, num_classes, memory_bank_index=0, temp=0.1,mask_memory=None):
b, num_feat, w, j = pred_rep.shape
device = pred_rep.device
pred_rep = pred_rep.permute(0, 2, 3, 1)
loss = torch.tensor(0.0).to(device)
class_count = 0
size = int(memory_bank.size(1)/2)
#memory_bank_l_mean=torch.mean(memory_bank[:,:size,:],dim=1)
#memory_bank_r_mean=torch.mean(memory_bank[:,size:,:],dim=1)
eps = 1e-12
for i in range(num_classes):
# print(pred_rep.shape)
per_count_size=0
with torch.no_grad():
negative_feat = memory_bank[[j for j in range(num_classes) if j != i]].view(-1, num_feat)
loss_pixel=torch.tensor(0.0).to(device)
which_choice=(mask & (labels == i))
if which_choice.sum()==0:
continue
class_count+=1
for which_net in range(2):
pred_rep_i = pred_rep[which_choice&(which_memory==(1-which_net))]
if len(pred_rep_i) == 0:
continue
per_count = len(pred_rep_i)
per_count_size+=per_count
with torch.no_grad():
if which_net==0:
positive_feat = memory_bank[i,:size,:]
#positive_feat=memory_bank_l_mean[i].unsqueeze(0)
else:
positive_feat=memory_bank[i,size:,:]
#positive_feat=memory_bank_r_mean[i].unsqueeze(0)
all_feats = torch.cat([positive_feat, negative_feat])
logits = ((pred_rep_i @ all_feats.T) / (torch.norm(pred_rep_i, dim=1).unsqueeze(0).T @ torch.norm(all_feats, dim=1).unsqueeze(0)))
logits = logits / temp
logits = torch.exp(logits)
logits_down = torch.sum(logits[:, size:], dim=1).unsqueeze(1)
logits_posi = logits[:, :size]
loss_pixel += (-torch.log((logits_posi / (logits_posi + logits_down + eps)) + eps)).sum()
loss += (loss_pixel/(size*per_count_size))
del mask,labels,which_memory
return loss / class_count
def contr_self_loss(memory_bank, pred_rep, labels, mask,num_classes, temp=0.1):
b, num_feat, w, j = pred_rep.shape
device = pred_rep.device
# 到这一步就有对应的labels和rep，每个像素都对应起来
pred_rep = pred_rep.permute(0, 2, 3, 1)
loss = torch.tensor(0.0).to(device)
class_count = 0
size = int(memory_bank.size(1))
#memory_bank_l_mean=torch.mean(memory_bank[:,:size,:],dim=1)
#memory_bank_r_mean=torch.mean(memory_bank[:,size:,:],dim=1)
eps = 1e-12
for i in range(num_classes):
# print(pred_rep.shape)
with torch.no_grad():
negative_feat = memory_bank[[j for j in range(num_classes) if j != i]].view(-1, num_feat)
positive_feat = memory_bank[i]
all_feats = torch.cat([positive_feat, negative_feat])
loss_pixel=torch.tensor(0.0).to(device)
which_choice=(mask & (labels == i))
if which_choice.sum()==0:
continue
class_count+=1
pred_rep_i = pred_rep[which_choice]
if len(pred_rep_i) == 0:
continue
per_count_size = len(pred_rep_i)
logits = ((pred_rep_i @ all_feats.T) / (torch.norm(pred_rep_i, dim=1).unsqueeze(0).T @ torch.norm(all_feats, dim=1).unsqueeze(0)))
logits = logits / temp
logits = torch.exp(logits)
logits_down = torch.sum(logits[:, size:], dim=1).unsqueeze(1)
logits_posi = logits[:, :size]
loss_pixel += (-torch.log((logits_posi / (logits_posi + logits_down + eps)) + eps)).sum()
loss += (loss_pixel/(size*per_count_size))
del mask,labels
return loss / class_count
def contr_net_loss(memory_bank, pred_rep_l, pred_rep_r, labels, mask, which_memory, num_classes,temp=0.5):
b, num_feat, w, j = pred_rep_l.shape
device = pred_rep_l.device
pred_rep_l = pred_rep_l.permute(0, 2, 3, 1)
pred_rep_r = pred_rep_r.permute(0, 2, 3, 1)
loss = torch.tensor(0.0).to(device)
class_count = 0
eps = 1e-12
for i in range(num_classes):
# print(pred_rep.shape)
per_count_size = 0
with torch.no_grad():
negative_feat = memory_bank[[j for j in range(num_classes) if j != i]].view(-1, num_feat)
#loss_pixel = torch.tensor(0.0).to(device)
which_choice = (mask & (labels == i))
if which_choice.sum() == 0:
continue
class_count += 1
pred_rep_i_l = pred_rep_l[which_choice & which_memory]
pred_rep_i_r = pred_rep_r[which_choice & which_memory]
per_count_size = len(pred_rep_i_l)
logits = ((pred_rep_i_l @ negative_feat.T) / (
torch.norm(pred_rep_i_l, dim=1).unsqueeze(0).T @ torch.norm(negative_feat, dim=1).unsqueeze(0)))
logits = logits / temp
logits = torch.exp(logits)
logits_down = torch.sum(logits, dim=1)
logits_posi=torch.sum(pred_rep_i_l*pred_rep_i_r,dim=1)/(torch.norm(pred_rep_i_l,dim=1)*torch.norm(pred_rep_i_r,dim=1))
logits_posi = logits_posi / temp
logits_posi = torch.exp(logits_posi)
loss_pixel = (-torch.log((logits_posi / (logits_posi + logits_down + eps)) + eps)).sum()
#print(per_count_size)
#print(logits.shape)
#print(logits_posi.shape)
#print(logits_down.shape)
loss += (loss_pixel / (per_count_size))
return loss / class_count
def main():
print(config)
best_mIoU_l = 0
best_mIoU_r = 0
if consistency_loss == 'CE':
if len(gpus) > 1:
unlabeled_loss = torch.nn.DataParallel(CrossEntropyLoss2dPixelWiseWeighted(ignore_index=ignore_label),device_ids=gpus).cuda()
#unlabeled_loss = torch.nn.DataParallel(DynamicMutualLoss(gamma1=2, gamma2=2, ignore_index=ignore_label),device_ids=gpus).cuda()
else:
unlabeled_loss = CrossEntropyLoss2dPixelWiseWeighted(ignore_index=ignore_label).cuda()
#unlabeled_loss = DynamicMutualLoss(gamma1=2, gamma2=2, ignore_index=ignore_label).cuda()
elif consistency_loss == 'MSE':
if len(gpus) > 1:
unlabeled_loss = torch.nn.DataParallel(MSELoss2d(), device_ids=gpus).cuda()
else:
unlabeled_loss = MSELoss2d().cuda()
cudnn.enabled = True
# create network
model_l = Res_Deeplab(num_classes=num_classes)
model_r = Res_Deeplab(num_classes=num_classes)
# load pretrained parameters
if restore_from_l[:4] == 'http':
saved_state_dict_l = model_zoo.load_url(restore_from_l)
saved_state_dict_r = model_zoo.load_url(restore_from_r)
else:
saved_state_dict_l = torch.load(restore_from_l)
saved_state_dict_r = model_zoo.load(restore_from_r)
# Copy loaded parameters to model
new_params = model_l.state_dict().copy()
for name, param in new_params.items():
if name in saved_state_dict_l and param.size() == saved_state_dict_l[name].size():
new_params[name].copy_(saved_state_dict_l[name])
model_l.load_state_dict(new_params)
new_params = model_r.state_dict().copy()
for name, param in new_params.items():
if name in saved_state_dict_r and param.size() == saved_state_dict_r[name].size():
new_params[name].copy_(saved_state_dict_r[name])
model_r.load_state_dict(new_params)
# Initiate ema-model
if train_unlabeled:
ema_model_l = create_ema_model(model_l)
ema_model_l.train()
ema_model_l = ema_model_l.cuda()
ema_model_r = create_ema_model(model_r)
ema_model_r.train()
ema_model_r = ema_model_r.cuda()
else:
ema_model_l = None
ema_model_r = None
if len(gpus) > 1:
if use_sync_batchnorm:
model_l = convert_model(model_l)
model_l = DataParallelWithCallback(model_l, device_ids=gpus)
model_r = convert_model(model_r)
model_r = DataParallelWithCallback(model_r, device_ids=gpus)
else:
model_l = torch.nn.DataParallel(model_l, device_ids=gpus)
model_r = torch.nn.DataParallel(model_r, device_ids=gpus)
model_l.train()
model_l.cuda()
model_r.train()
model_r.cuda()
cudnn.benchmark = True
if dataset == 'pascal_voc':
data_loader = get_loader(dataset)
data_path = get_data_path(dataset)
train_dataset = data_loader(data_path, crop_size=input_size, scale=random_scale, mirror=random_flip)
elif dataset == 'cityscapes':
data_loader = get_loader('cityscapes')
data_path = get_data_path('cityscapes')
if random_crop:
data_aug = RandomCrop_city(input_size)
else:
data_aug = None
train_dataset = data_loader(data_path, is_transform=True, augmentations=data_aug)
train_dataset_size = len(train_dataset)
print('dataset size: ', train_dataset_size)
partial_size = labeled_samples
print('Training on number of samples:', partial_size)
if split_id is not None:
train_ids = pickle.load(open(split_id, 'rb'))
print('loading train ids from {}'.format(split_id))
else:
np.random.seed(random_seed)
train_ids = np.arange(train_dataset_size)
np.random.shuffle(train_ids)
train_sampler = data.sampler.SubsetRandomSampler(train_ids[:partial_size])
trainloader = data.DataLoader(train_dataset,
batch_size=batch_size, sampler=train_sampler, num_workers=num_workers,
pin_memory=True)
trainloader_iter = iter(trainloader)
if train_unlabeled:
train_remain_sampler = data.sampler.SubsetRandomSampler(train_ids[partial_size:])
trainloader_remain = data.DataLoader(train_dataset,
batch_size=batch_size, sampler=train_remain_sampler, num_workers=num_workers,
pin_memory=True)
trainloader_remain_iter = iter(trainloader_remain)
# Optimizer for segmentation network
learning_rate_object = Learning_Rate_Object(config['training']['learning_rate'])
if optimizer_type == 'SGD':
if len(gpus) > 1:
optimizer_l = optim.SGD(model_l.module.optim_parameters(learning_rate_object),
lr=learning_rate, momentum=momentum, weight_decay=weight_decay)
optimizer_r = optim.SGD(model_r.module.optim_parameters(learning_rate_object),
lr=learning_rate, momentum=momentum, weight_decay=weight_decay)
else:
optimizer_l = optim.SGD(model_l.optim_parameters(learning_rate_object),
lr=learning_rate, momentum=momentum, weight_decay=weight_decay)
optimizer_r = optim.SGD(model_r.optim_parameters(learning_rate_object),
lr=learning_rate, momentum=momentum, weight_decay=weight_decay)
#model_l, optimizer_l = amp.initialize(model_l, optimizer_l, opt_level='O1')
#model_r, optimizer_r = amp.initialize(model_r, optimizer_r, opt_level='O1')
#model_l = DataParallelWithCallback(model_l, device_ids=gpus)
#model_r = DataParallelWithCallback(model_r, device_ids=gpus)
optimizer_l.zero_grad()
optimizer_r.zero_grad()
interp = nn.Upsample(size=(input_size[0], input_size[1]), mode='bilinear', align_corners=True)
start_iteration = 0
if args.resume:
start_iteration, model_l, optimizer_l, ema_model_l = _resume_checkpoint(args.resume, model_l, optimizer_l,
ema_model_l)
start_iteration, model_r, optimizer_r, ema_model_r = _resume_checkpoint(args.resume, model_r, optimizer_r,
ema_model_r)
accumulated_loss_l_l = []
accumulated_loss_l_r = []
if train_unlabeled:
accumulated_loss_u_l = []
accumulated_loss_u_r = []
if not os.path.exists(checkpoint_dir):
os.makedirs(checkpoint_dir)
with open(checkpoint_dir + '/config.json', 'w') as handle:
json.dump(config, handle, indent=4, sort_keys=False)
pickle.dump(train_ids, open(os.path.join(checkpoint_dir, 'train_split.pkl'), 'wb'))
class_sampling = [1.0 for i in range(num_classes)]
epochs_since_start = 0
with_embed = False
epoch = 0
epochs = math.ceil(num_iterations / (labeled_samples // batch_size))
loss_l_value_l = 0
loss_l_value_r = 0
if train_unlabeled:
loss_u_value_l = 0
loss_u_value_r = 0
total = 0
acc_l = 0
acc_r = 0
meaniou_l = 0
meaniou_r = 0
start_epoch = datetime.datetime.now()
end_epoch = datetime.datetime.now()
memory_bank_l = torch.randn(num_classes, every_memory_count, 256).cuda()
memory_bank_size_l = torch.zeros(num_classes).int()
memory_bank_r = torch.randn(num_classes, every_memory_count, 256).cuda()
memory_bank_size_r = torch.zeros(num_classes).int()
contr_train=True
for i_iter in range(start_iteration, num_iterations):
model_l.train()
model_r.train()
start = datetime.datetime.now()
optimizer_l.zero_grad()
optimizer_r.zero_grad()
if lr_schedule:
adjust_learning_rate(optimizer_l, i_iter)
adjust_learning_rate(optimizer_r, i_iter)
# Training loss for labeled data only
try:
batch = next(trainloader_iter)
if batch[0].shape[0] != batch_size:
batch = next(trainloader_iter)
except:
epochs_since_start = epochs_since_start + 1
# print('Epochs since start: ',epochs_since_start)
trainloader_iter = iter(trainloader)
batch = next(trainloader_iter)
weak_parameters = {"flip": 0}
images, labels, _, _, n = batch
images = images.cuda()
labels = labels.cuda()
images, labels = weakTransform(weak_parameters, data=images, target=labels)
pred_l = model_l(images.detach())
pred_r = model_r(images.detach())
L_l_l = loss_calc(interp(pred_l['out']), labels)
L_l_r = loss_calc(interp(pred_r['out']), labels)
labels_feat = F.interpolate(labels.clone().unsqueeze(1).float(), size=pred_l['feat'].shape[2:], mode='nearest').squeeze(1).long()
if train_unlabeled:
try:
batch_remain = next(trainloader_remain_iter)
if batch_remain[0].shape[0] != batch_size:
batch_remain = next(trainloader_remain_iter)
except:
trainloader_remain_iter = iter(trainloader_remain)
batch_remain = next(trainloader_remain_iter)
images_remain, _, _, _, _ = batch_remain
images_remain = images_remain.cuda()
inputs_u_w, _ = weakTransform(weak_parameters, data=images_remain)
# ema
L_u_ema_l,pred_u_l,prob_l,max_l = ema_loss(model_l, ema_model_r, images_remain, inputs_u_w, weak_parameters, interp, i_iter,unlabeled_loss)
L_u_ema_r,pred_u_r,prob_r,max_r = ema_loss(model_r, ema_model_l, images_remain, inputs_u_w, weak_parameters, interp, i_iter,unlabeled_loss)
#L_u_ema_l,L_u_ema_r = dual_teacher(model_l,model_r, ema_model_l,ema_model_r, images_remain, inputs_u_w, weak_parameters, interp, i_iter,unlabeled_loss)
# L_u_ema_l=torch.tensor(0.0)
# L_u_ema_r=torch.tensor(0.0)
if with_embed and contr_train==True:
gamma = 0.75
#pred_u_l=model_l(inputs_u_w.detach())
#pred_u_r=model_r(inputs_u_w.detach())
# pred_u_l={'out':torch.randn(2,19,33,65).cuda(),'feat':torch.randn(2,256,33,65).cuda()}
# pred_u_r = {'out':torch.randn(2, 19, 33, 65).cuda(),'feat':torch.randn(2,256,33,65).cuda()}
with torch.no_grad():
prob_u_l, max_u_l = torch.max(torch.softmax(pred_u_l['out'],dim=1), dim=1)
prob_u_r, max_u_r = torch.max(torch.softmax(pred_u_r['out'],dim=1), dim=1)
labels_l = torch.cat([labels_feat, max_l])
labels_r = torch.cat([labels_feat, max_r])
mask_l = torch.cat([(labels_feat != ignore_label), (prob_l >= gamma)])
mask_r = torch.cat([(labels_feat != ignore_label), (prob_r >= gamma)])
#prob_l=torch.cat([torch.max(torch.softmax(pred_l['out'],dim=1),dim=1)[0],prob_l])
#prob_r=torch.cat([torch.max(torch.softmax(pred_r['out'],dim=1),dim=1)[0],prob_r])
# mask_l[mask_l==False]=True
# mask_r[mask_r == False] = True
#which_memory=(prob_l>=prob_r).int()
#which_memory = (prob_u_l >= prob_u_r).int()
which_memory=torch.ones_like(prob_u_l).cuda()
pred_rep_l = torch.cat([pred_l['feat'], pred_u_l['feat']])
pred_rep_r = torch.cat([pred_r['feat'], pred_u_r['feat']])
mask_label=torch.zeros(labels_feat.shape).bool().cuda()
L_l_contr=contr_self_loss(memory_bank_l,pred_rep_l,labels_l,mask_l,num_classes)
L_r_contr=contr_self_loss(memory_bank_r,pred_rep_r,labels_r,mask_r,num_classes)
L_l_contr_cross = contr_net_loss(torch.cat([memory_bank_l, memory_bank_r], dim=1), pred_rep_l[batch_size:,:,:],pred_rep_r[batch_size:,:,:].detach(), max_u_r, mask_l[batch_size:,:,:].int()==0,
(which_memory==1) & (prob_u_r>=0), num_classes)
L_r_contr_cross = contr_net_loss(torch.cat([memory_bank_l, memory_bank_r], dim=1), pred_rep_r[batch_size:,:,:],pred_rep_l[batch_size:,:,:].detach(), max_u_l, mask_r[batch_size:,:,:].int()==0,
(which_memory==1) & (prob_u_l>=0), num_classes)
del mask_l,mask_r
if with_embed and contr_train==True:
loss_l = L_l_l +L_u_ema_l
loss_r = L_l_r +L_u_ema_r
else:
loss_l = L_l_l+L_u_ema_l
loss_r = L_l_r+L_u_ema_r
else:
loss_l = L_l_l
loss_r = L_l_r