-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathtrain_patch.py
1578 lines (1256 loc) · 69.2 KB
/
train_patch.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 os
import time
import math
import random
import numpy as np
import cv2
import h5py
from glob import glob
from scipy.ndimage.filters import gaussian_filter
from scipy.ndimage import sobel
import torch
import torch.backends.cudnn as cudnn
import torch.nn as nn
import torch.nn.functional as F
from torch import optim
from utils import *
from models import *
import ChamferDistancePytorch.chamfer3D.dist_chamfer_3D
chamLoss = ChamferDistancePytorch.chamfer3D.dist_chamfer_3D.chamfer_3DDist()
class MODEL_PATCH(object):
def __init__(self, config):
self.real_size = 256
#self.real_size = 128
self.mask_margin = 8
self.g_dim = config.g_dim
self.z_dim = config.z_dim
self.input_size = config.input_size # 32
self.output_size = config.output_size # 128
self.patch_size = config.patch_size
self.csize = config.csize
self.c_range = config.c_range
self.stride = 4
self.pt_num = 768
self.K = config.K
self.max_sample_num = config.max_sample_num
self.w_ident = config.w_ident
# self.up_rate = self.output_size//self.input_size
self.upsample_rate = 1
self.asymmetry = True
self.save_epoch = 5
self.eval_epoch = 2
self.start_epoch = 0
self.mode = config.mode
self.gua_filter = True
self.small_dataset = config.small_dataset
self.dump_deform =config.dump_deform
self.top_nums = [8]
self.pre_fix = ''
self.pred_coarse = True
self.dis_thres= 0.035
self.sampling_threshold = 0.3
self.pt_sampling_threshold = 0.3
self.render_view_id = 0
self.voxel_renderer = voxel_renderer(self.real_size)
self.checkpoint_dir = config.checkpoint_dir
self.data_dir = config.data_dir
self.dump_deform_path = config.dump_deform_path
if self.dump_deform and not os.path.exists(self.dump_deform_path):
os.makedirs(self.dump_deform_path)
self.data_content = config.data_content # content_chair
self.log_path = os.path.join(config.log_dir, self.model_dir, config.model_name)
self.sample_dir = os.path.join(config.sample_dir, self.model_dir, config.model_name)
if not os.path.exists(self.log_path):
os.makedirs(self.log_path)
if not os.path.exists(self.sample_dir):
os.makedirs(self.sample_dir)
self.log_fout = open(os.path.join(self.log_path, 'log_train.txt'), 'a')
self.log_fout.write(str(config)+'\n')
if torch.cuda.is_available():
self.device = torch.device('cuda')
torch.backends.cudnn.benchmark = True
else:
self.device = torch.device('cpu')
x_ = range(self.patch_size)
y_ = range(self.patch_size)
z_ = range(self.patch_size)
yy, xx, zz = np.meshgrid(x_,y_,z_)
self.patch_coord = np.concatenate((np.expand_dims(xx, -1), np.expand_dims(yy, -1),np.expand_dims(zz, -1)), -1)
#load data
print("preprocessing - start")
self.imgout_0 = np.full([self.real_size*4, self.real_size*4*3], 255, np.uint8)
if os.path.exists("splits/"+self.data_content+"_train.txt"):
#load data
fin = open("splits/"+self.data_content+"_%s.txt"%(self.mode))
self.dataset_names = [name.strip() for name in fin.readlines()]
fin.close()
self.dataset_len = len(self.dataset_names)
self.dataset_len = 1000
if self.mode=='test':
self.dataset_len = 200
if self.dump_deform and self.mode=='train':
fin2 = open("splits/"+self.data_content+"_test.txt")
test_names = [name.strip() for name in fin2.readlines()]
fin2.close()
self.dataset_names = self.dataset_names + test_names[:200]
self.dataset_len += 200
if self.small_dataset:
self.dataset_len = 4
self.mask_content = []
self.input_content = []
self.partial_content = []
self.partial_mask = []
self.gt_content = []
self.pos_content = []
self.plane_param = []
self.crop_locs=[]
self.pcn_partial = []
self.pcn_gt = []
self.gt_pc = []
self.shape_names = []
self.partial_or = []
self.pcn_pt_ids = []
self.pcn_rand_ids = []
# self.im_content = []
self.names = range(self.dataset_len)
vox_name = "/model.binvox"
if config.train_patch:
for i in range(self.dataset_len):
print("preprocessing content - "+str(i+1)+"/"+str(self.dataset_len))
if not os.path.exists(os.path.join(self.data_dir,self.dataset_names[self.names[i]]+vox_name)):
print('non exists')
self.dataset_len = self.dataset_len-1
continue
tmp_raw = get_vox_from_binvox_1over2(os.path.join(self.data_dir,self.dataset_names[self.names[i]]+vox_name)).astype(np.uint8)
xmin,xmax,ymin,ymax,zmin,zmax = self.get_voxel_bbox(tmp_raw)
tmp = self.crop_voxel(tmp_raw,xmin,xmax,ymin,ymax,zmin,zmax)
gt_voxel = tmp
tmp_input, tmp_mask = self.get_voxel_input_Dmask_mask(tmp)
if self.mode=='test' or self.dump_deform : prob=1
else: prob = None
partial_shape, partial_mask, c_locs = self.random_crop(tmp, crop_size=self.csize, \
c_range=self.c_range, prob=prob,return_locs=True)
partial_mask = tmp_mask
self.crop_locs.append(c_locs)
self.shape_names.append(self.dataset_names[self.names[i]])
if self.gua_filter:
tmp_input = gaussian_filter(tmp_input.astype(np.float32), sigma=1)
partial_shape = gaussian_filter(partial_shape.astype(np.float32), sigma=1)
self.input_content.append(tmp_input)
self.gt_content.append(gt_voxel)
self.mask_content.append(tmp_mask)
self.partial_content.append(partial_shape)
self.partial_mask.append(partial_mask)
self.pos_content.append([xmin,xmax,ymin,ymax,zmin,zmax])
img_y = i//4
img_x = (i%4)*3+2
if img_y<4:
# xmin,xmax,ymin,ymax,zmin,zmax = self.pos_content[i]
tmpvox = self.recover_voxel(gt_voxel,xmin,xmax,ymin,ymax,zmin,zmax)
self.imgout_0[img_y*self.real_size:(img_y+1)*self.real_size,img_x*self.real_size:(img_x+1)*self.real_size] = self.voxel_renderer.render_img(tmpvox, self.sampling_threshold, self.render_view_id)
img_y = i//4
img_x = (i%4)*3
if img_y<4:
tmpvox = self.recover_voxel(tmp_input,xmin,xmax,ymin,ymax,zmin,zmax)
self.imgout_0[img_y*self.real_size:(img_y+1)*self.real_size,img_x*self.real_size:(img_x+1)*self.real_size] = self.voxel_renderer.render_img(tmpvox, self.sampling_threshold, self.render_view_id)
img_y = i//4
img_x = (i%4)*3 + 1
if img_y<4:
tmpvox = self.recover_voxel(partial_shape,xmin,xmax,ymin,ymax,zmin,zmax)
self.imgout_0[img_y*self.real_size:(img_y+1)*self.real_size,img_x*self.real_size:(img_x+1)*self.real_size] = self.voxel_renderer.render_img(tmpvox, self.sampling_threshold, self.render_view_id)
cv2.imwrite(self.sample_dir+"/a_content_%s.png"%(self.mode), self.imgout_0)
else:
print("ERROR: cannot load dataset txt: "+"splits/"+self.data_content+"_train.txt")
exit(-1)
self.c_encoder = PatchEncoder(self.g_dim, self.z_dim)
self.d_encoder = PatchEncoder(self.g_dim, self.z_dim)
self.c_encoder.to(self.device)
self.d_encoder.to(self.device)
c_params = list(self.c_encoder.parameters())
d_params = list(self.d_encoder.parameters())
self.optimizer = torch.optim.Adam(c_params+d_params, lr=config.lr)
self.scheduler = torch.optim.lr_scheduler.StepLR(self.optimizer, 30, gamma=0.99)
if self.pred_coarse:
print('coarse predictor', self.data_content)
if self.data_content =='content_chair':
self.dis_thres = 0.035 # 0.055
checkpoint = torch.load('checkpoint_complete/content_chair_complete/coarse_comp/checkpoint-90.pth')
elif self.data_content=='content_boat':
self.dis_thres=0.12
checkpoint = torch.load('checkpoint_complete/content_boat_complete/coarse_comp/checkpoint-90.pth')
elif self.data_content=='content_lamp':
self.dis_thres=0.04
checkpoint = torch.load('checkpoint_complete/content_lamp_complete/coarse_comp/checkpoint-90.pth')
elif self.data_content=='content_couch':
self.dis_thres=0.12
checkpoint = torch.load('checkpoint_complete/content_couch_complete/coarse_comp/checkpoint-150.pth')
elif self.data_content=='content_cabinet':
self.dis_thres=0.045
checkpoint = torch.load('checkpoint_complete/content_cabinet_complete/coarse_comp/checkpoint-90.pth')
elif self.data_content=='content_plane':
self.dis_thres = 0.045
checkpoint = torch.load('checkpoint_complete/content_plane_complete/coarse_comp/checkpoint-90.pth')
elif self.data_content=='content_car':
self.dis_thres=0.05
checkpoint = torch.load('checkpoint_complete/content_car_complete/coarse_comp/checkpoint-90.pth') #
elif self.data_content=='content_table':
self.dis_thres = 0.045
checkpoint = torch.load('checkpoint_complete/content_table_complete/coarse_comp/checkpoint-90.pth')
self.coarse_predictor = CoarseCompletor_skip(32)
self.coarse_predictor.to(self.device)
self.coarse_predictor.load_state_dict(checkpoint['generator'])
self.coarse_predictor.eval()
self.max_to_keep = 3
self.checkpoint_path = os.path.join(self.checkpoint_dir, self.model_dir, config.model_name)
self.checkpoint_name='checkpoint'
self.checkpoint_manager_list = [None] * self.max_to_keep
self.checkpoint_manager_pointer = 0
def log_string(self, out_str, ):
self.log_fout.write(out_str+'\n')
self.log_fout.flush()
print(out_str)
def get_voxel_input_Dmask_mask(self,vox):
vox_tensor = torch.from_numpy(vox).to(self.device).unsqueeze(0).unsqueeze(0).float()
#input
down_rate = self.output_size//self.input_size
if self.output_size==128: pool_rate=3
else: pool_rate = 5
if down_rate==2:
smallmaskx_tensor = F.max_pool3d(vox_tensor, kernel_size = down_rate, stride = down_rate, padding = 0)
input_vox = F.interpolate(smallmaskx_tensor, scale_factor = down_rate, mode='nearest')
else:
input_vox = F.max_pool3d(vox_tensor, kernel_size = down_rate, stride = 1, padding = down_rate//2)
input_vox = input_vox[:,:,:-1,:-1,:-1]
#Dmask
# smallmask_tensor = vox_tensor[:,:,crop_margin:-crop_margin,crop_margin:-crop_margin,crop_margin:-crop_margin]
smallmask_tensor = F.max_pool3d(input_vox, kernel_size = self.patch_size, stride = self.stride, padding = 0)
#to numpy
input_in = input_vox.detach().cpu().numpy()[0,0]
input_in = np.round(input_in).astype(np.uint8)
mask_in = smallmask_tensor.detach().cpu().numpy()[0,0]
mask_in = np.round(mask_in).astype(np.uint8)
return input_in, mask_in
def random_crop(self, vox, crop_size=20, c_range=20, prob=None, return_locs=False ):
vx, vy, vz = vox.shape
edges = sobel(vox)
edges = edges.astype(np.float32)/255.0
csize = crop_size + (np.random.rand(3)-0.5)*c_range
csize[0] = min(vx-16, csize[0])
csize[1] = min(vy-16, csize[1])
csize[2] = min(vz-16, csize[2])
csize = csize.astype(np.int32)
loc_starts = np.zeros(3,np.int32)
new_vox = np.zeros(vox.shape)
new_vox[:,:,:] = vox[:,:,:].copy()
loc_starts[0] = np.random.randint(0, vx-csize[0])
loc_starts[1] = np.random.randint(0, vy-csize[1])
loc_starts[2] = np.random.randint(0, vz-csize[2])
if np.array(edges>0.4,np.int32).sum()<100:
p = 0
else:
if prob is None:
p = np.random.random()
else:
p = prob
if p<0.4 or not self.data_content=='content_chair':
cnt=0
while(vox[loc_starts[0]:loc_starts[0]+csize[0], loc_starts[1]:loc_starts[1]+csize[1],\
loc_starts[2]:loc_starts[2]+csize[2],].sum()<30 ):
csize = crop_size + (np.random.rand(3)-0.5)*c_range
csize[0] = min(vx-16, csize[0])
csize[1] = min(vy-16, csize[1])
csize[2] = min(vz-16, csize[2])
csize = csize.astype(np.int32)
loc_starts[0] = np.random.randint(0, vx-csize[0])
loc_starts[1] = np.random.randint(0, vy-csize[1])
loc_starts[2] = np.random.randint(0, vz-csize[2])
cnt+=1
if cnt>10:
break
else:
cnt=0
while(vox[loc_starts[0]:loc_starts[0]+csize[0], loc_starts[1]:loc_starts[1]+csize[1],\
loc_starts[2]:loc_starts[2]+csize[2],].sum()<50 or loc_starts[1]<vy/2):
csize = crop_size + (np.random.rand(3)-0.5)*c_range
csize[0] = min(vx-16, csize[0])
csize[1] = min(vy-16, csize[1])
csize[2] = min(vz-16, csize[2])
csize = csize.astype(np.int32)
loc_starts[0] = np.random.randint(0, vx-csize[0])
loc_starts[1] = np.random.randint(0, vy-csize[1])
loc_starts[2] = np.random.randint(0, vz-csize[2])
cnt+=1
if cnt>10:
break
new_vox[loc_starts[0]:loc_starts[0]+csize[0], loc_starts[1]:loc_starts[1]+csize[1],\
loc_starts[2]:loc_starts[2]+csize[2],] = 0
mask_in = np.zeros(vox.shape)
mask_in[loc_starts[0]:loc_starts[0]+csize[0], loc_starts[1]:loc_starts[1]+csize[1],\
loc_starts[2]:loc_starts[2]+csize[2],] = 1.0
if return_locs==True:
c_locs = np.zeros(6, np.int32)
c_locs[0:3] = loc_starts
c_locs[3:] = loc_starts+csize
return new_vox, mask_in,c_locs
else:
return new_vox, mask_in
def point_cloud_to_volume(self, points, vsize=256, radius=0.5):
""" input is Nx3 points.
output is vsize*vsize*vsize
assumes points are in range [-radius, radius]
"""
vol = np.zeros((vsize,vsize,vsize))
voxel = 2*radius/float(vsize)
locations = (points + radius)/voxel
locations = np.clip(locations, 0, vsize-1e-4)
locations = locations.astype(int)
vol[locations[:,0],locations[:,1],locations[:,2]] = 1.0
return vol
def get_voxel_bbox(self,vox):
#minimap
vox_tensor = torch.from_numpy(vox).to(self.device).unsqueeze(0).unsqueeze(0).float()
smallmaskx_tensor = F.max_pool3d(vox_tensor, kernel_size = self.upsample_rate, stride = self.upsample_rate, padding = 0)
smallmaskx_tensor = F.interpolate(smallmaskx_tensor, scale_factor = self.upsample_rate, mode='nearest')
smallmaskx = smallmaskx_tensor.detach().cpu().numpy()[0,0]
smallmaskx = np.round(smallmaskx).astype(np.uint8)
smallx,smally,smallz = smallmaskx.shape
#x
ray = np.max(smallmaskx,(1,2))
xmin = 0
xmax = 0
for i in range(smallx):
if ray[i]>0:
if xmin==0:
xmin = i
xmax = i
#y
ray = np.max(smallmaskx,(0,2))
ymin = 0
ymax = 0
for i in range(smally):
if ray[i]>0:
if ymin==0:
ymin = i
ymax = i
#z
ray = np.max(smallmaskx,(0,1))
if self.asymmetry:
zmin = 0
zmax = 0
for i in range(smallz):
if ray[i]>0:
if zmin==0:
zmin = i
zmax = i
else:
zmin = smallz//2
zmax = 0
for i in range(zmin,smallz):
if ray[i]>0:
zmax = i
return xmin,xmax+1,ymin,ymax+1,zmin,zmax+1
def crop_voxel(self,vox,xmin,xmax,ymin,ymax,zmin,zmax):
xspan = xmax-xmin
yspan = ymax-ymin
zspan = zmax-zmin
tmp = np.zeros([xspan*self.upsample_rate+self.mask_margin*2,yspan*self.upsample_rate+self.mask_margin*2,zspan*self.upsample_rate+self.mask_margin*2], np.uint8)
if self.asymmetry:
tmp[self.mask_margin:-self.mask_margin,self.mask_margin:-self.mask_margin,self.mask_margin:-self.mask_margin] = vox[xmin*self.upsample_rate:xmax*self.upsample_rate,ymin*self.upsample_rate:ymax*self.upsample_rate,zmin*self.upsample_rate:zmax*self.upsample_rate]
else:
#note z is special: only get half of the shape in z: 0 0.5-----1
tmp[self.mask_margin:-self.mask_margin,self.mask_margin:-self.mask_margin,:-self.mask_margin] = vox[xmin*self.upsample_rate:xmax*self.upsample_rate,ymin*self.upsample_rate:ymax*self.upsample_rate,zmin*self.upsample_rate-self.mask_margin:zmax*self.upsample_rate]
return tmp
def recover_voxel(self,vox,xmin,xmax,ymin,ymax,zmin,zmax):
tmpvox = np.zeros([self.real_size,self.real_size,self.real_size], np.float32)
xmin_,ymin_,zmin_ = (0,0,0)
xmax_,ymax_,zmax_ = vox.shape
xmin = xmin*self.upsample_rate-self.mask_margin
xmax = xmax*self.upsample_rate+self.mask_margin
ymin = ymin*self.upsample_rate-self.mask_margin
ymax = ymax*self.upsample_rate+self.mask_margin
if self.asymmetry:
zmin = zmin*self.upsample_rate-self.mask_margin
else:
zmin = zmin*self.upsample_rate
zmin_ = self.mask_margin
zmax = zmax*self.upsample_rate+self.mask_margin
if xmin<0:
xmin_ = -xmin
xmin = 0
if xmax>self.real_size:
xmax_ = xmax_+self.real_size-xmax
xmax = self.real_size
if ymin<0:
ymin_ = -ymin
ymin = 0
if ymax>self.real_size:
ymax_ = ymax_+self.real_size-ymax
ymax = self.real_size
if zmin<0:
zmin_ = -zmin
zmin = 0
if zmax>self.real_size:
zmax_ = zmax_+self.real_size-zmax
zmax = self.real_size
if self.asymmetry:
tmpvox[xmin:xmax,ymin:ymax,zmin:zmax] = vox[xmin_:xmax_,ymin_:ymax_,zmin_:zmax_]
else:
tmpvox[xmin:xmax,ymin:ymax,zmin:zmax] = vox[xmin_:xmax_,ymin_:ymax_,zmin_:zmax_]
if zmin*2-zmax-1<0:
tmpvox[xmin:xmax,ymin:ymax,zmin-1::-1] = vox[xmin_:xmax_,ymin_:ymax_,zmin_:zmax_]
else:
tmpvox[xmin:xmax,ymin:ymax,zmin-1:zmin*2-zmax-1:-1] = vox[xmin_:xmax_,ymin_:ymax_,zmin_:zmax_]
return tmpvox
def pad_to_same_size(self, v1, v2):
_,_,x1,y1,z1 = v1.shape
_,_,x2,y2,z2 = v2.shape
x = max(x1, x2)
y = max(y1, y2)
z = max(z1, z2)
new_v1 = F.pad(v1, (int((z-z1)/2), z-z1-int((z-z1)/2), int((y-y1)/2), y-y1-int((y-y1)/2), int((x-x1)/2), x-x1-int((x-x1)/2)))
new_v2 = F.pad(v2, (int((z-z2)/2), z-z2-int((z-z2)/2), int((y-y2)/2), y-y2-int((y-y2)/2), int((x-x2)/2), x-x2-int((x-x2)/2)))
return new_v1, new_v2
def reshape_to_patch_size(self, v1, v2):
x1,y1,z1 = v1.shape
x,y,z = v2.shape
new_v1 = v1
padding = np.zeros(6)
if z1 < z:
padding[0], padding[1] = int((z-z1)/2), z-z1-int((z-z1)/2)
else:
new_v1 = new_v1[:, :,int((z1-z)/2):int((z1-z)/2)+z ]
if y1 < y:
padding[2], padding[3] = int((y-y1)/2), y-y1-int((y-y1)/2)
else:
new_v1 = new_v1[:, int((y1-y)/2):int((y1-y)/2)+y,: ]
if x1 < x:
padding[4], padding[5] = int((x-x1)/2), x-x1-int((x-x1)/2)
else:
new_v1 = new_v1[int((x1-x)/2):int((x1-x)/2)+x,:,: ]
new_v1 = torch.from_numpy(new_v1,).unsqueeze(0).unsqueeze(0)
new_v1 = F.pad(new_v1, tuple(padding.astype(np.int8)))
return new_v1.numpy()[0,0]
def reshape_to_size_torch(self, v1, v2_shape):
_,_,x1,y1,z1 = v1.shape
x,y,z = v2_shape
new_v1 = v1
padding = np.zeros(6)
if z1 < z:
padding[0], padding[1] = int((z-z1)/2), z-z1-int((z-z1)/2)
else:
new_v1 = new_v1[:,:,:, :,int((z1-z)/2):int((z1-z)/2)+z ]
if y1 < y:
padding[2], padding[3] = int((y-y1)/2), y-y1-int((y-y1)/2)
else:
new_v1 = new_v1[:,:,:, int((y1-y)/2):int((y1-y)/2)+y,: ]
if x1 < x:
padding[4], padding[5] = int((x-x1)/2), x-x1-int((x-x1)/2)
else:
new_v1 = new_v1[:,:,int((x1-x)/2):int((x1-x)/2)+x,:,: ]
new_v1 = F.pad(new_v1, tuple(padding.astype(np.int8)))
return new_v1
def get_exact_partial_mask(self, partial_in, gt_in, p_mask_in):
partial_in = (partial_in>0.3).float()
vox_dif = torch.abs(gt_in - partial_in)
patch_sum = F.avg_pool3d(partial_in, kernel_size = self.patch_size, stride = self.stride, padding = 0)
patch_sum = patch_sum*float(self.patch_size**3)
patch_dif = F.avg_pool3d(vox_dif, kernel_size = self.patch_size, stride = self.stride, padding = 0)
patch_dif = patch_dif*float(self.patch_size**3)
#patch_dif = (patch_dif<220).float()
patch_dif = ((patch_dif/(patch_sum+1e-3))<0.8).float()
mask_final = p_mask_in*patch_dif
return mask_final
def load(self):
#load previous checkpoint
checkpoint_txt = os.path.join(self.checkpoint_path, "checkpoint")
print(self.checkpoint_path)
if os.path.exists(checkpoint_txt):
fin = open(checkpoint_txt)
model_dir = fin.readline().strip()
print('loading from', model_dir)
fin.close()
checkpoint = torch.load(model_dir)
self.c_encoder.load_state_dict(checkpoint['c_encoder'])
self.d_encoder.load_state_dict(checkpoint['d_encoder'])
if 'optimizer' in checkpoint.keys():
self.optimizer.load_state_dict(checkpoint['optimizer'])
self.scheduler.load_state_dict(checkpoint['scheduler'])
self.start_epoch = checkpoint['epoch'] + 1
print(" [*] Load SUCCESS")
return True
else:
print(" [!] Load failed...")
return False
def save(self,epoch):
if not os.path.exists(self.checkpoint_path):
os.makedirs(self.checkpoint_path)
save_dir = os.path.join(self.checkpoint_path,self.checkpoint_name+"-"+str(epoch)+".pth")
self.checkpoint_manager_pointer = (self.checkpoint_manager_pointer+1)%self.max_to_keep
#delete checkpoint
if self.checkpoint_manager_list[self.checkpoint_manager_pointer] is not None:
if os.path.exists(self.checkpoint_manager_list[self.checkpoint_manager_pointer]):
os.remove(self.checkpoint_manager_list[self.checkpoint_manager_pointer])
#save checkpoint
torch.save({
'c_encoder': self.c_encoder.state_dict(),
'd_encoder': self.d_encoder.state_dict(),
'optimizer': self.optimizer.state_dict(),
'scheduler': self.scheduler.state_dict(),
'epoch': epoch
}, save_dir)
# update checkpoint manager
self.checkpoint_manager_list[self.checkpoint_manager_pointer] = save_dir
#write file
checkpoint_txt = os.path.join(self.checkpoint_path, "checkpoint")
fout = open(checkpoint_txt, 'w')
for i in range(self.max_to_keep):
pointer = (self.checkpoint_manager_pointer+self.max_to_keep-i)%self.max_to_keep
if self.checkpoint_manager_list[pointer] is not None:
fout.write(self.checkpoint_manager_list[pointer]+"\n")
fout.close()
def filter_feat(self, feat, mask):
_, _, x_size, y_size, z_size = mask.shape
x_ = range(x_size)
y_ = range(y_size)
z_ = range(z_size)
feats = feat[0].permute(1,2,3,0)
yy, xx, zz = np.meshgrid(y_,x_,z_)
feat_coord = np.concatenate((np.expand_dims(xx, -1), np.expand_dims(yy, -1),np.expand_dims(zz, -1)), -1)
filtered_feat = feats[mask[0,0]>0]
filtered_coord = feat_coord[mask[0,0].cpu().numpy()>0]
return filtered_feat,filtered_coord
def distance_retrieve(self, c_feat, d_feat, K=20,sigma=1.0, c_max_num=200, d_max_num=200):
"""
feat: [1, nf, z_dim]
return:
c_ids, d_ids, distance matrix
"""
c_num = min(c_feat.shape[0], c_max_num)
d_num = min(d_feat.shape[0], d_max_num)
c_ids = np.array(sorted(np.random.randint(0, c_feat.shape[0], c_num)))
d_ids = np.array(sorted(np.random.randint(0, d_feat.shape[0], d_num)))
c_feat_chosen = c_feat[c_ids,:].detach().cpu()
d_feat_chosen = d_feat[d_ids,:].detach().cpu()
c_feat_chosen = c_feat_chosen.view(c_num, 1, -1)
d_feat_chosen = d_feat_chosen.view(1, d_num, -1)
dis_matrix = torch.sqrt(torch.sum(torch.square(c_feat_chosen - d_feat_chosen), -1)) #[c_num, d_num]
dis_prob = dis_matrix / sigma
dis_prob = torch.exp(-(dis_prob - torch.max(dis_prob, dim=1).values))
dis_prob = dis_prob.numpy()
dis_prob = dis_prob/(dis_prob.sum(1)+1e-7)
#dis_prob[:,0] = dis_prob[:,0] + 1- dis_prob.sum(1)
#dis_prob[:,0] = dis_prob[:,0] + 1- dis_prob.sum(1)
out_ids = np.zeros((c_num, K))
for i in range(c_num):
if dis_prob[i].sum()<1e-5:
out_ids[i] = np.random.randint(0, d_num, size=K)
continue
if not dis_prob[i].sum()<1000:
out_ids[i] = np.random.randint(0, d_num, size=K)
continue
if dis_prob[i].sum()!=1:
dis_prob[i] = dis_prob[i]/dis_prob[i].sum()
#print(dis_prob[i].sum())
tmp_d_ids = np.random.choice(np.arange(d_num), size=K, p=dis_prob[i])
tmp_d_ids = tmp_d_ids.astype(np.int32)
out_ids[i] = d_ids[tmp_d_ids]
return c_ids, out_ids.astype(np.int32)
def random_retrieve(self,num_c, num_d, K=20, c_max_num=200,):
c_num = min(num_c, c_max_num)
c_ids = sorted(np.random.randint(0, num_c, c_num))
d_ids = np.zeros((c_num, K))
for i in range(c_num):
d_ids[i] = np.random.randint(0,num_d, size=K)
return c_ids, d_ids
def compute_feat_dis(self,c_feats_less, d_feats_less, re_c_ids, re_d_ids):
c_feats_less = c_feats_less[re_c_ids]
out_dis = torch.zeros(re_d_ids.shape).to(self.device)
for i in range(re_d_ids.shape[1]):
d_feat_tmp = d_feats_less[re_d_ids[:,i]]
out_dis[:,i] = torch.sqrt(torch.sum(torch.square(c_feats_less - d_feat_tmp), -1))
return out_dis
def regression_loss(self, embedding_distance, actual_distance, mask=None,obj_sigmas=1.0):
if mask is not None:
loss = torch.sum(torch.square(mask*(embedding_distance-actual_distance)),)/(mask.sum()+1e-4)
else:
loss = torch.mean(torch.square(embedding_distance-actual_distance),)
return loss
def compute_iou(self, recons, gt, thresh=0.4):
x = (recons>thresh).astype(np.float32)
y = (gt>thresh).astype(np.float32)
intersection = np.sum(x*y)
union = np.sum(x) + np.sum(y) - np.sum(x*y)
iou = intersection/(union+1e-5)
return iou
def get_ids_with_gt(self, c_locs, total_d_locs):
c_num = c_locs.shape[0]
valid_ids = []
for i in range(c_num):
dif = np.abs(total_d_locs - c_locs[i,]).sum(1).min()
if dif==0:
valid_ids.append(i)
return valid_ids
@property
def model_dir(self):
return "{}_retriv".format(self.data_content)
def train(self, config):
if self.dump_deform:
self.mode = 'test'
if config.continue_train or self.mode=='test':
self.load()
start_time = time.time()
training_epoch = config.epoch
if self.mode=='test':
training_epoch += 1
batch_index_list = np.arange(self.dataset_len)
for epoch in range(self.start_epoch, training_epoch):
if self.mode!='test':
np.random.shuffle(batch_index_list)
total_loss = 0.0
total_iou = 0.0
total_c_iou = 0.0
total_cd = 0.0
self.c_encoder.train()
self.d_encoder.train()
if self.mode=='test':
self.c_encoder.eval()
self.d_encoder.eval()
print('model eval..')
#self.dataset_len = len(self.input_content)
for idx in range(self.dataset_len):
dxb = batch_index_list[idx]
qxp = dxb
if not self.pred_coarse:
input_in = torch.from_numpy(self.input_content[dxb]).to(self.device).unsqueeze(0).unsqueeze(0).float()
mask_in = torch.from_numpy(self.mask_content[dxb]).to(self.device).unsqueeze(0).unsqueeze(0).float()
gt_in = torch.from_numpy(self.gt_content[dxb]).to(self.device).unsqueeze(0).unsqueeze(0).float()
if self.mode=='test':
partial_shape = self.partial_content[qxp]
partial_mask = self.partial_mask[qxp]
else:
partial_shape, partial_mask = self.random_crop(self.gt_content[dxb],crop_size=self.csize, \
c_range=self.c_range)
partial_mask = self.mask_content[dxb]
if self.gua_filter:
partial_shape = gaussian_filter(partial_shape.astype(np.float32), sigma=1)
partial_in = torch.from_numpy(partial_shape).to(self.device).unsqueeze(0).unsqueeze(0).float()
mask_partial_in = torch.from_numpy(partial_mask).to(self.device).unsqueeze(0).unsqueeze(0).float()
if self.pred_coarse:
if self.mode=='test':
partial_tmp = self.partial_content[dxb]
else:
partial_tmp, partial_mask_tmp = self.random_crop(self.gt_content[dxb],crop_size=self.csize, \
c_range=self.c_range)
if self.gua_filter:
partial_tmp = gaussian_filter(partial_tmp.astype(np.float32), sigma=1)
partial_tmp_in = torch.from_numpy(partial_tmp).to(self.device).unsqueeze(0).unsqueeze(0).float()
partial_mask_2 = F.max_pool3d((partial_tmp_in>self.sampling_threshold).float(), \
kernel_size = self.patch_size, stride = self.stride, padding = 0)
partial_full = self.reshape_to_size_torch(partial_tmp_in, (self.output_size,self.output_size,self.output_size))
partial_in = partial_tmp_in
mask_partial_in = partial_mask_2
coarse_out = self.coarse_predictor(partial_full)
coarse_out = coarse_out.detach()
input_in = self.reshape_to_size_torch(coarse_out, partial_shape.shape)
mask_in = F.max_pool3d((input_in>self.sampling_threshold).float(), kernel_size = self.patch_size, \
stride = self.stride, padding = 0)
self.optimizer.zero_grad()
c_feat = self.c_encoder(input_in,is_training=True)
d_feat = self.d_encoder(partial_in,is_training=True)
# 1. choose coarse, detailed pairs, compute feature distance
# 2. compute gt distance
# 3. compute loss
c_feats_less, c_locs = self.filter_feat(c_feat, mask_in)
d_feats_less, d_locs = self.filter_feat(d_feat, mask_partial_in)
for top_num in self.top_nums:
if self.dump_deform and (epoch==self.start_epoch or epoch%20==0):
out_ids, out_dis, out_trans = self.retrieve_to_full_shape_topk(c_feats_less, \
c_locs, d_feats_less, d_locs, gt_in, partial_in,input_in, \
dxb, coarse=False, return_k=True, top_num=top_num, target_pt_num=self.pt_num)
out_ids_c, out_dis_c, out_trans_c = self.retrieve_to_full_shape_topk(c_feats_less, c_locs, \
d_feats_less, d_locs, gt_in, partial_in,input_in, dxb,coarse=True, return_k=True,\
top_num=top_num, sigma=1, target_pt_num=self.pt_num)
out_path = os.path.join(self.dump_deform_path, self.dataset_names[self.names[dxb]],)
crop_locs = self.crop_locs[dxb]
if not os.path.exists(out_path):
os.makedirs(out_path)
if self.pred_coarse:
coarse_shape = input_in.cpu().numpy()[0,0]
else:
coarse_shape = 0
np.savez(os.path.join(out_path, 'dump_deform_wc_top%d_%d.npz'%(top_num, epoch)), re_ids=out_ids, \
re_dis=out_dis, re_trans=out_trans, re_ids_c=out_ids_c, re_dis_c=out_dis_c, \
re_trans_c=out_trans_c, \
coarse_shape=coarse_shape, crop_locs=crop_locs)
if dxb>20 and self.dataset_len>50:
continue
if self.dump_deform:
out_path_im = self.dump_deform_path
else:
out_path_im = self.sample_dir
if self.mode=='test' and (epoch==self.start_epoch or epoch%20==0):
if self.pred_coarse:
imgout_0 = np.full([self.real_size, self.real_size*7], 255, np.uint8)
else:
imgout_0 = np.full([self.real_size, self.real_size*5], 255, np.uint8)
xmin,xmax,ymin,ymax,zmin,zmax = self.pos_content[dxb]
if self.pred_coarse:
tmpvox = self.recover_voxel(partial_shape, xmin,xmax,ymin,ymax,zmin,zmax)
imgout_0[:,(-2)*self.real_size:-self.real_size] = self.voxel_renderer.render_img(tmpvox, \
self.sampling_threshold, self.render_view_id)
tmpvox = self.recover_voxel(input_in.cpu().numpy()[0][0], xmin,xmax,ymin,ymax,zmin,zmax)
imgout_0[:,-self.real_size:] = self.voxel_renderer.render_img(tmpvox, self.sampling_threshold, \
self.render_view_id)
tmpvox = self.recover_voxel(self.input_content[dxb],xmin,xmax,ymin,ymax,zmin,zmax)
imgout_0[:,0:self.real_size] = self.voxel_renderer.render_img(tmpvox, self.sampling_threshold, self.render_view_id)
tmpvox = self.recover_voxel(partial_in.cpu().numpy()[0][0],xmin,xmax,ymin,ymax,zmin,zmax)
imgout_0[:,self.real_size:self.real_size*2] = self.voxel_renderer.render_img(tmpvox, self.sampling_threshold,\
self.render_view_id)
tmpvox = self.recover_voxel(gt_in.cpu().numpy()[0][0],xmin,xmax,ymin,ymax,zmin,zmax)
imgout_0[:,4*self.real_size:self.real_size*5] = self.voxel_renderer.render_img(tmpvox, self.sampling_threshold,\
self.render_view_id)
out_vox, out_vox_or, dis = self.retrieve_to_full_shape_topk(c_feats_less, c_locs, d_feats_less,\
d_locs, gt_in, partial_in,input_in,dxb,top_num=top_num, \
apply_mask=True,target_pt_num=self.pt_num,)
iou = self.compute_iou(out_vox, gt_in.cpu().numpy()[0][0],thresh=self.sampling_threshold)
total_iou += iou
print(idx ,iou, total_iou/(idx+1))
out_vox = self.reshape_to_patch_size(out_vox, partial_shape)
out_vox_or = self.reshape_to_patch_size(out_vox_or, partial_shape)
tmpvox = self.recover_voxel(out_vox,xmin,xmax,ymin,ymax,zmin,zmax)
imgout_0[:,3*self.real_size:4*self.real_size] = self.voxel_renderer.render_img(tmpvox, \
self.sampling_threshold, self.render_view_id)
tmpvox = self.recover_voxel(out_vox_or,xmin,xmax,ymin,ymax,zmin,zmax)
imgout_0[:,2*self.real_size:3*self.real_size] = self.voxel_renderer.render_img(tmpvox, \
self.sampling_threshold, self.render_view_id)
if self.dump_deform:
cv2.imwrite(self.dump_deform_path+"/top%d_%d_%d_%s_%.3f.png"%(top_num,epoch, self.names[dxb], \
self.pre_fix, iou), imgout_0)
else:
cv2.imwrite(self.sample_dir+"/top%d_%s_%d_%d_%.3f.png"%(top_num, self.pre_fix, epoch, \
self.names[dxb], iou), imgout_0)
if self.mode=='test' and not config.continue_train:
continue
num_c = c_feats_less.shape[0]
num_d = d_feats_less.shape[0]
if epoch < 5000:
try:
re_c_ids, re_d_ids = self.random_retrieve(num_c, num_d, K=self.K, c_max_num=self.max_sample_num) # [300], [300, 20]
except:
continue
else:
re_c_ids, re_d_ids, = self.distance_retrieve(c_feats_less, d_feats_less, K=self.K)
re_dis = self.compute_feat_dis(c_feats_less, d_feats_less, re_c_ids, re_d_ids)
gt_dis = self.get_icp_dis(gt_in, partial_in, input_in, c_locs, d_locs, re_c_ids,re_d_ids, \
target_pt_num=self.pt_num, check=False)
# print("gt_dis", gt_dis, gt_dis.mean(), gt_dis.min(), gt_dis.max())
gt_dis = gt_dis*(gt_dis<self.dis_thres) + 10*gt_dis*(gt_dis>=self.dis_thres)
gt_dis = torch.from_numpy(gt_dis).float().to(self.device)
same_d_ids = np.arange(d_locs.shape[0]).reshape(-1, 1)
same_c_ids = np.zeros(same_d_ids.shape[0])
same_mask = np.ones(d_locs.shape[0])
for i in range(same_d_ids.shape[0]):
loc_dif = np.sum(np.square((c_locs - d_locs[i])), axis=1)
if loc_dif.min()>0:
same_mask[i]=0
continue
same_c_ids[i] = np.where(loc_dif==0)[0][0]
same_mask = same_mask.astype(np.bool)
same_c_ids = same_c_ids.astype(np.int32)
same_c_ids = same_c_ids[same_mask]
same_d_ids = same_d_ids[same_mask]
same_gt_dis = self.get_icp_dis(gt_in, partial_in,input_in, c_locs, d_locs, same_c_ids, same_d_ids, \
target_pt_num=self.pt_num, check=False)
#print('same_gt_dis', same_gt_dis, same_gt_dis.mean(), same_gt_dis.min(), same_gt_dis.max())
same_re_dis = self.compute_feat_dis(c_feats_less, d_feats_less, same_c_ids, same_d_ids)
same_gt_dis = torch.from_numpy(same_gt_dis).float().to(self.device)
# print('same_gt_dis', same_gt_dis, same_gt_dis.mean(), same_gt_dis.min(), same_gt_dis.max())
same_mask = (same_gt_dis<self.dis_thres).float().to(self.device)
loss_identity = self.regression_loss(same_re_dis, same_gt_dis, mask=same_mask)
loss_rand = self.regression_loss(re_dis, gt_dis)
# print(dxb,partial_in.sum(), loss_rand.item(),loss_identity.item())
loss = loss_rand + self.w_ident*loss_identity
loss.backward()
self.optimizer.step()
if self.mode=='test':
return
self.log_string("Epoch: [%d/%d] time: %.0f, loss_rand: %.5f, loss_itent: %.5f" % \
(epoch, training_epoch, time.time() - start_time, loss_rand.item(),loss_identity.item(), ))
if epoch % 5==0 and self.mode=='train':
imgout_0 = np.full([self.real_size, self.real_size*5], 255, np.uint8)
xmin,xmax,ymin,ymax,zmin,zmax = self.pos_content[dxb]
if self.pred_coarse:
imgout_0 = np.full([self.real_size, self.real_size*7], 255, np.uint8)
tmpvox = self.recover_voxel(partial_tmp, xmin,xmax,ymin,ymax,zmin,zmax)
imgout_0[:,5*self.real_size:6*self.real_size] = self.voxel_renderer.render_img(tmpvox, self.sampling_threshold, \
self.render_view_id)
tmpvox = self.recover_voxel(input_in.cpu().numpy()[0][0], xmin,xmax,ymin,ymax,zmin,zmax)
imgout_0[:,6*self.real_size:7*self.real_size] = self.voxel_renderer.render_img(tmpvox, self.sampling_threshold, \
self.render_view_id)
tmpvox = self.recover_voxel(input_in.cpu().numpy()[0,0],xmin,xmax,ymin,ymax,zmin,zmax)
imgout_0[:,0:self.real_size] = self.voxel_renderer.render_img(tmpvox, self.sampling_threshold, self.render_view_id)
tmpvox = self.recover_voxel(partial_in.cpu().numpy()[0][0],xmin,xmax,ymin,ymax,zmin,zmax)
imgout_0[:,self.real_size:self.real_size*2] = self.voxel_renderer.render_img(tmpvox, self.sampling_threshold, \
self.render_view_id)
tmpvox = self.recover_voxel(gt_in.cpu().numpy()[0,0],xmin,xmax,ymin,ymax,zmin,zmax)
imgout_0[:,4*self.real_size:self.real_size*5] = self.voxel_renderer.render_img(tmpvox, self.sampling_threshold, \
self.render_view_id)
out_vox, out_vox_or, dis = \
self.retrieve_to_full_shape(c_feats_less, c_locs, d_feats_less, d_locs, gt_in, partial_in,input_in, dxb,\
target_pt_num=self.pt_num)
tmpvox = self.recover_voxel(out_vox,xmin,xmax,ymin,ymax,zmin,zmax)
imgout_0[:,3*self.real_size:4*self.real_size] = self.voxel_renderer.render_img(tmpvox, self.sampling_threshold, \
self.render_view_id)
tmpvox = self.recover_voxel(out_vox_or,xmin,xmax,ymin,ymax,zmin,zmax)