forked from HiLab-git/IPLC
-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
741 lines (651 loc) · 26.3 KB
/
utils.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
##load data
from torch.utils.data import Dataset
import numpy as np
import os
import torch
import yaml
import random
import configparser
import math
from torch.nn import init
from scipy import ndimage
import SimpleITK as sitk
import torch.nn as nn
import cv2
import pandas as pd
import matplotlib.pyplot as plt
def random_sample(LV):
LV[LV == 1] = 255
LV = LV.astype(np.uint8)
white_indices_LV = np.where(LV == 255)
points = []
input_labels = []
process_part_1(white_indices_LV, points, input_labels)
if all(points_part is None for points_part in points):
return None
else:
all_points = np.concatenate([points_part for points_part in points if points_part is not None])
all_input_labels = np.concatenate([input_labels_part for input_labels_part in input_labels if input_labels_part is not None])
return all_points, all_input_labels
def process_part_center(MYO, points, input_labels):
contours, _ = cv2.findContours(MYO, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
if len(contours) == 0:
points.append(None)
input_labels.append(None)
else:
largest_contour = max(contours, key=cv2.contourArea)
contour = largest_contour
M = cv2.moments(contour)
if M['m00'] != 0:
centroid_x = int(M['m10'] / M['m00'])
centroid_y = int(M['m01'] / M['m00'])
center_point = (centroid_x, centroid_y)
points.append(np.array([center_point]))
input_labels.append(np.array([1]))
def process_part_center_0(MYO, points, input_labels):
contours, _ = cv2.findContours(MYO, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
if len(contours) == 0:
points.append(None)
input_labels.append(None)
else:
largest_contour = max(contours, key=cv2.contourArea)
contour = largest_contour
M = cv2.moments(contour)
if M['m00'] != 0:
centroid_x = int(M['m10'] / M['m00'])
centroid_y = int(M['m01'] / M['m00'])
center_point = (centroid_x, centroid_y)
points.append(np.array([center_point]))
input_labels.append(np.array([0]))
def process_part_1(white_indices, points, input_labels):
if white_indices[0].size < 3:
points.append(None)
input_labels.append(None)
else:
random_indices = np.random.choice(len(white_indices[0]), size=3, replace=False)
random_points = [(white_indices[0][i], white_indices[1][i]) for i in random_indices]
points.append(np.array(random_points))
points[-1] = np.column_stack((points[-1][:, 1], points[-1][:, 0]))
input_labels.append(np.array([1, 1, 1]))
def process_part_0(white_indices, points, input_labels):
if white_indices[0].size < 3:
points.append(None)
input_labels.append(None)
else:
random_indices = np.random.choice(len(white_indices[0]), size=3, replace=False)
random_points = [(white_indices[0][i], white_indices[1][i]) for i in random_indices]
points.append(np.array(random_points))
points[-1] = np.column_stack((points[-1][:, 1], points[-1][:, 0]))
input_labels.append(np.array([0, 0, 0]))
def norm_0_255(x):
x_shape = list(x.shape)
if (len(x_shape) == 5):
[N, C, D, H, W] = x_shape
new_shape = [N * D, C, H, W]
x = torch.transpose(x, 1, 2)
x = torch.reshape(x, new_shape)
normalized_x = ((x + 1) / 2) * 255
normalized_x = normalized_x.to(torch.uint8)
image_array = normalized_x.cpu().detach().numpy()
return image_array
def show_mask(mask, img, random_color=False):
if random_color:
color = np.concatenate([np.random.random(3), np.array([0.6])], axis=0)
else:
color = np.array([30 / 255, 144 / 255, 255 / 255, 0.3])
h, w = mask.shape[-2:]
mask_image = mask.reshape(h, w, 1) * color.reshape(1, 1, -1)
mask_image_bgr = cv2.cvtColor(mask_image, cv2.COLOR_RGBA2BGRA) if mask_image.shape[2] == 4 else mask_image
img_bgr = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR) if len(img.shape) == 2 else img.copy()
combined_img = cv2.addWeighted(img_bgr, 1, mask_image_bgr, 0.7, 0)
return combined_img
def show_mask_and_points(mask, points, labels, img, output, gt):
marker_size = 2
pos_points = points[labels == 1]
neg_points = points[labels == 0]
color_sam = np.array([30, 30, 150, 80])
color_output = np.array([150, 30, 30, 80])
color_gt = np.array([30 / 255, 150 / 255, 30 / 255, 80 / 255])
h, w = mask.shape[-2:]
mask_image_sam = mask.reshape(h, w, 1)
mask_image_sam_1 = mask_image_sam.copy().astype(np.uint8)
contours_sam, _ = cv2.findContours(mask_image_sam_1.copy(), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
mask_image_output = output.reshape(h, w, 1)
mask_image_output_1 = mask_image_output.copy().astype(np.uint8)
contours_output, _ = cv2.findContours(mask_image_output_1.copy(), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
mask_image_gt = gt.reshape(h, w, 1)
mask_image_gt_1 = mask_image_gt.copy().astype(np.uint8)
contours_gt, _ = cv2.findContours(mask_image_gt_1.copy(), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
img_bgr = img
combined_img = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2BGRA)
cv2.drawContours(combined_img, contours_sam, -1, (0, 0, 255, 255), thickness=1)
cv2.drawContours(combined_img, contours_output, -1, (255, 0, 0, 255), thickness=1)
cv2.drawContours(combined_img, contours_gt, -1, (0, 255, 0, 255), thickness=1)
for pos_point in pos_points:
cv2.circle(combined_img, pos_point, marker_size, color=(0, 255, 0, 255), thickness=-1)
for neg_point in neg_points:
cv2.circle(combined_img, neg_point, marker_size, color=(0, 0, 255, 255), thickness=-1)
return combined_img
def show_points(coords, labels, img, marker_size=2):
pos_points = coords[labels == 1]
neg_points = coords[labels == 0]
image = img.copy()
image[image == 1] = 255
# Convert image to BGR format if not already in that format
img_bgr = cv2.cvtColor(image, cv2.COLOR_GRAY2BGR) if len(image.shape) == 2 else image.copy()
# Draw green and red points on the image
for pos_point in pos_points:
cv2.drawMarker(img_bgr, tuple(map(int, pos_point)), color=(0, 255, 0), markerType=cv2.MARKER_STAR,
markerSize=marker_size, thickness=1)
for neg_point in neg_points:
cv2.drawMarker(img_bgr, tuple(map(int, neg_point)), color=(0, 0, 255), markerType=cv2.MARKER_STAR,
markerSize=marker_size, thickness=1)
return img_bgr
def show_points_GT(coords, labels, img, marker_size=2):
pos_points = coords[labels == 1]
neg_points = coords[labels == 0]
image = img.copy()
# Convert image to BGR format if not already in that format
img_bgr = cv2.cvtColor(image, cv2.COLOR_GRAY2BGR) if len(image.shape) == 2 else image.copy()
# Draw green and red points on the image
for pos_point in pos_points:
cv2.drawMarker(img_bgr, tuple(map(int, pos_point)), color=(0, 255, 0), markerType=cv2.MARKER_STAR,
markerSize=marker_size, thickness=1)
for neg_point in neg_points:
cv2.drawMarker(img_bgr, tuple(map(int, neg_point)), color=(0, 0, 255), markerType=cv2.MARKER_STAR,
markerSize=marker_size, thickness=1)
return img_bgr
def is_int(val_str):
start_digit = 0
if(val_str[0] =='-'):
start_digit = 1
flag = True
for i in range(start_digit, len(val_str)):
if(str(val_str[i]) < '0' or str(val_str[i]) > '9'):
flag = False
break
return flag
def is_float(val_str):
flag = False
if('.' in val_str and len(val_str.split('.'))==2 and not('./' in val_str)):
if(is_int(val_str.split('.')[0]) and is_int(val_str.split('.')[1])):
flag = True
else:
flag = False
elif('e' in val_str and val_str[0] != 'e' and len(val_str.split('e'))==2):
if(is_int(val_str.split('e')[0]) and is_int(val_str.split('e')[1])):
flag = True
else:
flag = False
else:
flag = False
return flag
def is_bool(var_str):
if( var_str.lower() =='true' or var_str.lower() == 'false'):
return True
else:
return False
def parse_bool(var_str):
if(var_str.lower() =='true'):
return True
else:
return False
def is_list(val_str):
if(val_str[0] == '[' and val_str[-1] == ']'):
return True
else:
return False
def parse_list(val_str):
sub_str = val_str[1:-1]
splits = sub_str.split(',')
output = []
for item in splits:
item = item.strip()
if(is_int(item)):
output.append(int(item))
elif(is_float(item)):
output.append(float(item))
elif(is_bool(item)):
output.append(parse_bool(item))
elif(item.lower() == 'none'):
output.append(None)
else:
output.append(item)
return output
def parse_value_from_string(val_str):
if(is_int(val_str)):
val = int(val_str)
elif(is_float(val_str)):
val = float(val_str)
elif(is_list(val_str)):
val = parse_list(val_str)
elif(is_bool(val_str)):
val = parse_bool(val_str)
elif(val_str.lower() == 'none'):
val = None
else:
val = val_str
return val
def parse_config(filename):
config = configparser.ConfigParser()
config.read(filename)
output = {}
for section in config.sections():
output[section] = {}
for key in config[section]:
val_str = str(config[section][key])
if(len(val_str)>0):
val = parse_value_from_string(val_str)
output[section][key] = val
else:
val = None
print(section, key, val_str, val)
return output
def load_npz(path):
img = np.load(path)['arr_0']
gt = np.load(path)['arr_1']
return img, gt
def get_config(config):
with open(config, 'r') as stream:
return yaml.load(stream,Loader=yaml.FullLoader)
def set_random(seed_id=3407):
np.random.seed(seed_id)
torch.manual_seed(seed_id) #for cpu
torch.cuda.manual_seed_all(seed_id) #for GPU
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
# config setting
def is_int(val_str):
start_digit = 0
if(val_str[0] =='-'):
start_digit = 1
flag = True
for i in range(start_digit, len(val_str)):
if(str(val_str[i]) < '0' or str(val_str[i]) > '9'):
flag = False
break
return flag
def is_float(val_str):
flag = False
if('.' in val_str and len(val_str.split('.'))==2 and not('./' in val_str)):
if(is_int(val_str.split('.')[0]) and is_int(val_str.split('.')[1])):
flag = True
else:
flag = False
elif('e' in val_str and val_str[0] != 'e' and len(val_str.split('e'))==2):
if(is_int(val_str.split('e')[0]) and is_int(val_str.split('e')[1])):
flag = True
else:
flag = False
else:
flag = False
return flag
def is_bool(var_str):
if( var_str.lower() =='true' or var_str.lower() == 'false'):
return True
else:
return False
def parse_bool(var_str):
if(var_str.lower() =='true'):
return True
else:
return False
def is_list(val_str):
if(val_str[0] == '[' and val_str[-1] == ']'):
return True
else:
return False
def parse_list(val_str):
sub_str = val_str[1:-1]
splits = sub_str.split(',')
output = []
for item in splits:
item = item.strip()
if(is_int(item)):
output.append(int(item))
elif(is_float(item)):
output.append(float(item))
elif(is_bool(item)):
output.append(parse_bool(item))
elif(item.lower() == 'none'):
output.append(None)
else:
output.append(item)
return output
def parse_value_from_string(val_str):
if(is_int(val_str)):
val = int(val_str)
elif(is_float(val_str)):
val = float(val_str)
elif(is_list(val_str)):
val = parse_list(val_str)
elif(is_bool(val_str)):
val = parse_bool(val_str)
elif(val_str.lower() == 'none'):
val = None
else:
val = val_str
return val
def parse_config(filename):
config = configparser.ConfigParser()
config.read(filename)
output = {}
for section in config.sections():
output[section] = {}
for key in config[section]:
val_str = str(config[section][key])
if(len(val_str)>0):
val = parse_value_from_string(val_str)
output[section][key] = val
else:
val = None
print(section, key, val_str, val)
return output
class UnpairedDataset(Dataset):
#get unpaired dataset, such as MR-CT dataset
def __init__(self,A_path,B_path):
listA = os.listdir(A_path)
listB = os.listdir(B_path)
self.listA = [os.path.join(A_path,k) for k in listA]
self.listB = [os.path.join(B_path,k) for k in listB]
self.Asize = len(self.listA)
self.Bsize = len(self.listB)
self.dataset_size = max(self.Asize,self.Bsize)
def __getitem__(self,index):
if self.Asize == self.dataset_size:
A,A_gt = load_npz(self.listA[index])
B,B_gt = load_npz(self.listB[random.randint(0, self.Bsize - 1)])
else :
B,B_gt = load_npz(self.listB[index])
A,A_gt = load_npz(self.listA[random.randint(0, self.Asize - 1)])
A = torch.from_numpy(A.copy()).unsqueeze(0).float()
A_gt = torch.from_numpy(A_gt.copy()).unsqueeze(0).float()
B = torch.from_numpy(B.copy()).unsqueeze(0).float()
B_gt = torch.from_numpy(B_gt.copy()).unsqueeze(0).float()
return A,A_gt,B,B_gt
def __len__(self):
return self.dataset_size
def crop_depth(img,lab,phase = 'train'):
D,H,W = img.shape
if D > 10:
if phase == 'train':
# target_ssh = np.random.randint(0, int(D-10), 1)[0]
# zero_img = img[target_ssh:target_ssh+10,:,:]
# zero_lab = lab[target_ssh:target_ssh+10,:,:]
target_ssh = int((D-10)/2)
zero_img = img[target_ssh:target_ssh+10,:,:]
zero_lab = lab[target_ssh:target_ssh+10,:,:]
elif phase == 'valid':
zero_img,zero_lab = img,lab
elif phase == 'feta':
sample_indices = np.random.choice(D, size=10, replace=False)
zero_img = np.zeros((10,H,W))
zero_lab = np.zeros((10,H,W))
for i, index in enumerate(sample_indices):
zero_img[i] = img[index]
zero_lab[i] = lab[index]
else:
zero_img = np.zeros((10,H,W))
zero_lab = np.zeros((10,H,W))
zero_img[0:D,:,:] = img
zero_lab[0:D,:,:] = lab
return zero_img,zero_lab
def winadj_mri(array):
v0 = np.percentile(array, 1)
v1 = np.percentile(array, 99)
array[array < v0] = v0
array[array > v1] = v1
v0 = array.min()
v1 = array.max()
array = (array - v0) / (v1 - v0) * 2.0 - 1.0
return array
def resize(img,lab):
D,H,W = img.shape
zoom = [1,256/H,256/W]
img=ndimage.zoom(img,zoom,order=2)
lab=ndimage.zoom(lab,zoom,order=0)
return img,lab
class niiDataset(Dataset):
def __init__(self, source_img,source_lab,dataset,target, phase = 'test'):
self.dataset = dataset
self.source_img = source_img
self.source_lab = source_lab
self.phase = phase
nii_names = os.listdir(source_img)
self.all_files = []
for nii_name in nii_names:
self.img_path = os.path.join(self.source_img, nii_name)
if self.dataset == 'mms':
self.lab_path = os.path.join(self.source_lab, nii_name[:-7] + '_gt.nii.gz')
else:
print(self.dataset)
raise Exception('Unrecognized dataset.')
self.nii_name = str(nii_name)
self.all_files.append({
"img": self.img_path,
"lab": self.lab_path,
"img_name": self.nii_name
})
def __getitem__(self, index):
fname = self.all_files[index]
img_obj = sitk.ReadImage(fname["img"])
A = sitk.GetArrayFromImage(img_obj) /1
lab_obj = sitk.ReadImage(fname["lab"])
A_gt = sitk.GetArrayFromImage(lab_obj)
if self.phase == 'train' and self.dataset == 'mms':
A,A_gt = crop_depth(A,A_gt)
A = winadj_mri(A)
A,A_gt = resize(A,A_gt)
A = torch.from_numpy(A.copy()).unsqueeze(0).float()
A_gt = torch.from_numpy(A_gt.copy()).unsqueeze(0).float()
return A, A_gt, fname["img_name"],fname["lab"]
def __len__(self):
return len(self.all_files)
def init_weights(net, init_type='normal', init_gain=0.02):
def init_func(m): # define the initialization function
classname = m.__class__.__name__
if hasattr(m, 'weight') and (classname.find('Conv') != -1 or classname.find('Linear') != -1):
if init_type == 'normal':
init.normal_(m.weight.data, 0.0, init_gain)
elif init_type == 'xavier':
init.xavier_normal_(m.weight.data, gain=init_gain)
elif init_type == 'kaiming':
init.kaiming_normal_(m.weight.data, a=0, mode='fan_in')
elif init_type == 'orthogonal':
init.orthogonal_(m.weight.data, gain=init_gain)
else:
raise NotImplementedError('initialization method [%s] is not implemented' % init_type)
if hasattr(m, 'bias') and m.bias is not None:
init.constant_(m.bias.data, 0.0)
elif classname.find('BatchNorm2d') != -1: # BatchNorm Layer's weight is not a matrix; only normal distribution applies.
init.normal_(m.weight.data, 1.0, init_gain)
init.constant_(m.bias.data, 0.0)
print('initialize network with %s' % init_type)
net.apply(init_func) # apply the initialization function <init_func>
def one_hot_encode(input_tensor):
if len(input_tensor.shape) == 4:
a,b,c,d = input_tensor.shape
elif len(input_tensor.shape) == 3:
input_tensor = input_tensor.unsqueeze(1)
tensor_list = []
for i in range(4):
tmp = (input_tensor==i) * torch.ones_like(input_tensor)
tensor_list.append(tmp)
output_tensor = torch.cat(tensor_list,dim=1)
return output_tensor.float()
def get_largest_component(image):
dim = len(image.shape)
if(image.sum() == 0 ):
# print('the largest component is null')
return image
if(dim == 2):
s = ndimage.generate_binary_structure(2,1)
elif(dim == 3):
s = ndimage.generate_binary_structure(3,1)
else:
raise ValueError("the dimension number should be 2 or 3")
labeled_array, numpatches = ndimage.label(image, s)
sizes = ndimage.sum(image, labeled_array, range(1, numpatches + 1))
max_label = np.where(sizes == sizes.max())[0] + 1
output = np.asarray(labeled_array == max_label[0], np.uint8)
return output
def tensor_rot_90(x):
x_shape = list(x.shape)
if(len(x_shape) == 4):
return x.flip(3).transpose(2, 3)
else:
return x.flip(2).transpose(1, 2)
def tensor_rot_180(x):
x_shape = list(x.shape)
if(len(x_shape) == 4):
return x.flip(3).flip(2)
else:
return x.flip(2).flip(1)
def tensor_flip_2(x):
x_shape = list(x.shape)
if(len(x_shape) == 4):
return x.flip(2)
else:
return x.flip(1)
def tensor_flip_3(x):
x_shape = list(x.shape)
if(len(x_shape) == 4):
return x.flip(3)
else:
return x.flip(2)
def tensor_rot_270(x):
x_shape = list(x.shape)
if(len(x_shape) == 4):
return x.transpose(2, 3).flip(3)
else:
return x.transpose(1, 2).flip(2)
def rotate_single_random(img):
x_shape = list(img.shape)
if(len(x_shape) == 5):
[N, C, D, H, W] = x_shape
new_shape = [N*D, C, H, W]
x = torch.transpose(img, 1, 2)
img = torch.reshape(x, new_shape)
label = np.random.randint(0, 4, 1)[0]
if label == 1:
img = tensor_rot_90(img)
elif label == 2:
img = tensor_rot_180(img)
elif label == 3:
img = tensor_rot_270(img)
else:
img = img
return img,label
def rotate_single_with_label(img, label):
if label == 1:
img = tensor_rot_90(img)
elif label == 2:
img = tensor_rot_180(img)
elif label == 3:
img = tensor_rot_270(img)
else:
img = img
return img
def random_rotate(A,A_gt):
target_ssh = np.random.randint(0, 8, 1)[0]
A = rotate_single_with_label(A, target_ssh)
A_gt = rotate_single_with_label(A_gt, target_ssh)
return A,A_gt
def rotate_4(img):
# target_ssh = np.random.randint(0, 4, 1)[0]
A_1 = rotate_single_with_label(img, 1)
A_2 = rotate_single_with_label(img, 2)
A_3 = rotate_single_with_label(img, 3)
return A_1,A_2,A_3
def SAM_get_pl_label(images, output_dec1, predictor,
mask_input , num_classes, sample_times):
if (len(images.shape) == 5):
[N, C, D, H, W] = images.shape
new_shape = [N * D, C, H, W]
x = torch.transpose(images, 1, 2)
images = torch.reshape(x, new_shape)
times = sample_times
images_all_depth = norm_0_255(images)
all_depth_weight_entropy = np.zeros((output_dec1.shape[0], 256, 256),dtype=np.float64)
sam_pseudo_label = np.zeros((output_dec1.shape[0], num_classes, 256, 256), dtype=np.uint8)
for depth in range(output_dec1.shape[0]):
for c in range(num_classes):
output_dec1[depth, c, :, :] = get_largest_component(output_dec1[depth, c, :, :])
images_one_depth = np.zeros((256, 256, 3), dtype=np.uint8)
images_one_depth[:, :, 0] = images_all_depth[depth, 0, :, :]
images_one_depth[:, :, 1] = images_all_depth[depth, 0, :, :]
images_one_depth[:, :, 2] = images_all_depth[depth, 0, :, :]
predictor.set_image(images_one_depth.copy())
mask_input_this_depth = np.zeros((num_classes, 64, 64), dtype=np.float64)
for c in range(num_classes):
resized_mask = cv2.resize(mask_input[depth, c, :, :], (64, 64), interpolation=cv2.INTER_LINEAR)
mask_input_this_depth[c, :, :] = resized_mask
mask_input_this_depth = mask_input_this_depth[np.newaxis, :, :, :]
this_depth_mask_all = np.zeros((times, num_classes, 256, 256), dtype=np.float64)
this_depth_probability_map = np.zeros((num_classes, 256, 256), dtype=np.float64)
for i in range(times):
for c in range(num_classes-1):
result = random_sample(output_dec1[depth, c+1, :, :].copy())
if result is not None:
points, labels = result
masks, scores, logits = predictor.predict(
point_coords=points,
point_labels=labels,
mask_input=mask_input_this_depth[:, c+1, :, :],
multimask_output=True,
)
this_depth_mask_all[i, c+1, :, :] = logits.copy()
else:
masks, scores, logits = predictor.predict(
point_coords=None,
point_labels=None,
mask_input=mask_input_this_depth[:, c+1, :, :],
multimask_output=True,
)
this_depth_mask_all[i, c+1, :, :] = logits.copy()
Fore = np.sum(this_depth_mask_all[i, :, :, :], axis=0)
Back = 1 - Fore
Back[Back < 0] = 0
this_depth_mask_all[i, 0, :, :] = Back
for c in range(num_classes):
this_depth_probability_map[c, :, :] = np.sum(this_depth_mask_all[:, c, :, :], axis=0) / times
this_depth_entropy = -np.sum(this_depth_probability_map.copy() * np.log2(this_depth_probability_map.copy() + 1e-10), axis=0)
this_depth_weight_entropy = (2 - this_depth_entropy) / 2
all_depth_weight_entropy[depth, :, :] = this_depth_weight_entropy.copy()
this_depth_argmax_map = np.argmax(this_depth_probability_map.copy(), axis=0)
for c in range(num_classes):
sam_pseudo_label[depth, c, :, :] = (this_depth_argmax_map == c)
for i in range(num_classes):
sam_pseudo_label[depth, i, :, :] = get_largest_component(sam_pseudo_label[depth, i, :, :])
return sam_pseudo_label, all_depth_weight_entropy
def show_mask_1(mask, img, output, gt):
marker_size = 2
color_sam = np.array([30/255, 30/255, 150/255, 80/255])
color_output = np.array([150, 30, 30, 80])
color_gt = np.array([30 / 255, 150 / 255, 30 / 255, 80 / 255])
h, w = mask.shape[-2:]
mask_image_sam = mask.reshape(h, w, 1)
mask_image_sam_1 = mask_image_sam.copy().astype(np.uint8)
contours_sam, _ = cv2.findContours(mask_image_sam_1.copy(), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
mask_image_output = output.reshape(h, w, 1)
mask_image_output_1 = mask_image_output.copy().astype(np.uint8)
contours_output, _ = cv2.findContours(mask_image_output_1.copy(), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
mask_image_gt = gt.reshape(h, w, 1)
mask_image_gt_1 = mask_image_gt.copy().astype(np.uint8)
contours_gt, _ = cv2.findContours(mask_image_gt_1.copy(), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
mask_image_sam = mask_image_sam * color_sam.reshape(1, 1, -1)
mask_image_sam = mask_image_sam.astype(np.uint8)
mask_image_output = mask_image_output * color_output.reshape(1, 1, -1)
mask_image_output = mask_image_output.astype(np.uint8)
mask_image_gt = mask_image_gt * color_gt.reshape(1, 1, -1)
mask_image_gt = mask_image_gt.astype(np.uint8)
img_bgr = img
# img_bgr = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2BGRA)
combined_img = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2BGRA)
# combined_img = cv2.addWeighted(img_bgr, 1, mask_image_sam, 0.7, 0)
# combined_img = cv2.addWeighted(combined_img, 1, mask_image_output, 0.7, 0)
# combined_img = cv2.addWeighted(combined_img, 1, mask_image_gt, 0.7, 0)
cv2.drawContours(combined_img, contours_sam, -1, (0, 0, 255, 255), thickness=1)
cv2.drawContours(combined_img, contours_output, -1, (255, 0, 0, 255), thickness=1)
cv2.drawContours(combined_img, contours_gt, -1, (0, 255, 0, 255), thickness=1)
return combined_img