-
Notifications
You must be signed in to change notification settings - Fork 16
/
datasets.py
566 lines (420 loc) · 17.3 KB
/
datasets.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
import os
import os.path
import random
import numpy as np
from PIL import Image
import scipy.io as sio
import torch
import torch.utils.data as data
from torchvision import transforms
from torchvision.transforms import ToTensor
to_tensor = ToTensor()
def make_dataset(root):
return [(os.path.join(root, 'hazy', img_name),
os.path.join(root, 'trans', img_name),
os.path.join(root, 'gt', img_name))
for img_name in os.listdir(os.path.join(root, 'hazy'))]
def make_dataset_its(root):
items = []
for img_name in os.listdir(os.path.join(root, 'hazy')):
idx0, idx1, ato = os.path.splitext(img_name)[0].split('_')
gt = os.path.join(root, 'clear', idx0 + '.png')
trans = os.path.join(root, 'trans', idx0 + '_' + idx1 + '.png')
haze = os.path.join(root, 'hazy', img_name)
items.append([haze, trans, float(ato), gt])
return items
def make_dataset_ots(root):
items = []
for img_name in os.listdir(os.path.join(root, 'haze')):
idx, _, _ = os.path.splitext(img_name)[0].split('_')
gt = os.path.join(root, 'clear', idx + '.jpg')
haze = os.path.join(root, 'haze', img_name)
items.append([haze, gt])
return items
def make_dataset_ohaze(root: str, mode: str):
img_list = []
for img_name in os.listdir(os.path.join(root, mode, 'hazy')):
gt_name = img_name.replace('hazy', 'GT')
assert os.path.exists(os.path.join(root, mode, 'gt', gt_name))
img_list.append([os.path.join(root, mode, 'hazy', img_name),
os.path.join(root, mode, 'gt', gt_name)])
return img_list
def make_dataset_oihaze_train(root, suffix):
items = []
for img_name in os.listdir(os.path.join(root, 'haze' + suffix)):
gt = os.path.join(root, 'gt' + suffix, img_name)
haze = os.path.join(root, 'haze' + suffix, img_name)
items.append((haze, gt))
return items
def make_dataset_oihaze_train_triple(root, suffix):
items = []
for img_name in os.listdir(os.path.join(root, 'haze' + suffix)):
haze = os.path.join(root, 'haze' + suffix, img_name)
gt = os.path.join(root, 'gt' + suffix, img_name)
predict = os.path.join(root, 'predict' + suffix, img_name)
items.append((haze, gt, predict))
return items
def make_dataset_oihaze_test(root):
items = []
for img_name in os.listdir(os.path.join(root, 'haze')):
img_f_name, img_l_name = os.path.splitext(img_name)
gt_name = '%sGT%s' % (img_f_name[: -4], img_l_name)
gt = os.path.join(root, 'gt', gt_name)
haze = os.path.join(root, 'haze', img_name)
items.append((haze, gt))
return items
def random_crop(size, haze, gt, extra=None):
w, h = haze.size
assert haze.size == gt.size
if w < size or h < size:
haze = transforms.Resize(size)(haze)
gt = transforms.Resize(size)(gt)
w, h = haze.size
x1 = random.randint(0, w - size)
y1 = random.randint(0, h - size)
_haze = haze.crop((x1, y1, x1 + size, y1 + size))
_gt = gt.crop((x1, y1, x1 + size, y1 + size))
if extra is None:
return _haze, _gt
else:
# extra: trans or predict
assert haze.size == extra.size
_extra = extra.crop((x1, y1, x1 + size, y1 + size))
return _haze, _gt, _extra
class ImageFolder(data.Dataset):
def __init__(self, root, flip=False, crop=None):
self.root = root
self.imgs = make_dataset(root)
self.gt_ato_dict = sio.loadmat(os.path.join(root, 'ato.mat'))
self.flip = flip
self.crop = crop
def __getitem__(self, index):
haze_path, trans_path, gt_path = self.imgs[index]
name = os.path.splitext(os.path.split(haze_path)[1])[0]
haze = Image.open(haze_path).convert('RGB')
trans = Image.open(trans_path).convert('L')
gt = Image.open(gt_path).convert('RGB')
assert haze.size == trans.size
assert trans.size == gt.size
if self.crop:
haze, trans, gt = random_crop(self.crop, haze, trans, gt)
if self.flip and random.random() < 0.5:
haze = haze.transpose(Image.FLIP_LEFT_RIGHT)
trans = trans.transpose(Image.FLIP_LEFT_RIGHT)
gt = gt.transpose(Image.FLIP_LEFT_RIGHT)
haze = to_tensor(haze)
trans = to_tensor(trans)
gt = to_tensor(gt)
gt_ato = torch.Tensor([self.gt_ato_dict[name][0, 0]]).float()
return haze, trans, gt_ato, gt, name
def __len__(self):
return len(self.imgs)
class ItsDataset(data.Dataset):
"""
For RESIDE Indoor
"""
def __init__(self, root, flip=False, crop=None):
self.root = root
self.imgs = make_dataset_its(root)
self.flip = flip
self.crop = crop
def __getitem__(self, index):
haze_path, trans_path, ato, gt_path = self.imgs[index]
name = os.path.splitext(os.path.split(haze_path)[1])[0]
haze = Image.open(haze_path).convert('RGB')
trans = Image.open(trans_path).convert('L')
gt = Image.open(gt_path).convert('RGB')
assert haze.size == trans.size
assert trans.size == gt.size
if self.crop:
haze, gt, trans = random_crop(self.crop, haze, gt, trans)
if self.flip and random.random() < 0.5:
haze = haze.transpose(Image.FLIP_LEFT_RIGHT)
trans = trans.transpose(Image.FLIP_LEFT_RIGHT)
gt = gt.transpose(Image.FLIP_LEFT_RIGHT)
haze = to_tensor(haze)
trans = to_tensor(trans)
gt = to_tensor(gt)
gt_ato = torch.Tensor([ato]).float()
return haze, trans, gt_ato, gt, name
def __len__(self):
return len(self.imgs)
class OtsDataset(data.Dataset):
"""
For RESIDE Outdoor
"""
def __init__(self, root, flip=False, crop=None):
self.root = root
self.imgs = make_dataset_ots(root)
self.flip = flip
self.crop = crop
def __getitem__(self, index):
haze_path, gt_path = self.imgs[index]
name = os.path.splitext(os.path.split(haze_path)[1])[0]
haze = Image.open(haze_path).convert('RGB')
gt = Image.open(gt_path).convert('RGB')
assert haze.size == gt.size
if self.crop:
haze, gt = random_crop(self.crop, haze, gt)
if self.flip and random.random() < 0.5:
haze = haze.transpose(Image.FLIP_LEFT_RIGHT)
gt = gt.transpose(Image.FLIP_LEFT_RIGHT)
haze = to_tensor(haze)
gt = to_tensor(gt)
return haze, gt, name
def __len__(self):
return len(self.imgs)
class SotsDataset(data.Dataset):
def __init__(self, root, mode=None):
self.root = root
self.imgs = make_dataset(root)
self.mode = mode
def __getitem__(self, index):
haze_path, trans_path, gt_path = self.imgs[index]
name = os.path.splitext(os.path.split(haze_path)[1])[0]
haze = Image.open(haze_path).convert('RGB')
haze = to_tensor(haze)
idx0 = name.split('_')[0]
gt = Image.open(os.path.join(self.root, 'gt', idx0 + '.png')).convert('RGB')
gt = to_tensor(gt)
if gt.shape != haze.shape:
# crop the indoor images
gt = gt[:, 10: 470, 10: 630]
return haze, gt, name
def __len__(self):
return len(self.imgs)
class OHazeDataset(data.Dataset):
def __init__(self, root, mode):
self.root = root
self.mode = mode
self.imgs = make_dataset_ohaze(root, mode)
def __getitem__(self, index):
haze_path, gt_path = self.imgs[index]
name = os.path.splitext(os.path.split(haze_path)[1])[0]
img = Image.open(haze_path).convert('RGB')
gt = Image.open(gt_path).convert('RGB')
if 'train' in self.mode:
# img, gt = random_crop(416, img, gt)
if random.random() < 0.5:
img = img.transpose(Image.FLIP_LEFT_RIGHT)
gt = gt.transpose(Image.FLIP_LEFT_RIGHT)
rotate_degree = np.random.choice([-90, 0, 90, 180])
img, gt = img.rotate(rotate_degree, Image.BILINEAR), gt.rotate(rotate_degree, Image.BILINEAR)
return to_tensor(img), to_tensor(gt), name
def __len__(self):
return len(self.imgs)
class OIHaze(data.Dataset):
def __init__(self, root, mode, suffix=None, flip=False, crop=None):
assert mode in ['train', 'test']
self.root = root
self.mode = mode
if mode == 'train':
self.img_name_list = make_dataset_oihaze_train(root, suffix)
else:
self.img_name_list = make_dataset_oihaze_test(root)
self.flip = flip
self.crop = crop
def __getitem__(self, index):
haze_path, gt_path = self.img_name_list[index]
name = os.path.splitext(os.path.split(haze_path)[1])[0]
haze = Image.open(haze_path).convert('RGB')
gt = Image.open(gt_path).convert('RGB')
if self.crop:
haze, gt = random_crop(self.crop, haze, gt)
if self.flip and random.random() < 0.5:
haze = haze.transpose(Image.FLIP_LEFT_RIGHT)
gt = gt.transpose(Image.FLIP_LEFT_RIGHT)
haze = to_tensor(haze)
gt = to_tensor(gt)
return haze, gt, name
def __len__(self):
return len(self.img_name_list)
class OIHaze5(data.Dataset):
def __init__(self, root, flip=False, rotate=None, resize=1024):
self.root = root
self.img_name_list = make_dataset_oihaze_test(root)
self.flip = flip
self.rotate = rotate
self.resize = transforms.Resize(resize)
def __getitem__(self, index):
haze_path, gt_path = self.img_name_list[index]
name = os.path.splitext(os.path.split(haze_path)[1])[0]
haze = Image.open(haze_path).convert('RGB')
gt = Image.open(gt_path).convert('RGB')
if self.flip and random.random() < 0.5:
haze = haze.transpose(Image.FLIP_LEFT_RIGHT)
gt = gt.transpose(Image.FLIP_LEFT_RIGHT)
if self.rotate:
rotate_degree = random.random() * 2 * self.rotate - self.rotate
haze = haze.rotate(rotate_degree, Image.BILINEAR)
gt = gt.rotate(rotate_degree, Image.BILINEAR)
haze_resize, gt_resize = self.resize(haze), self.resize(gt)
haze, gt = to_tensor(haze), to_tensor(gt)
haze_resize, gt_resize = to_tensor(haze_resize), to_tensor(gt_resize)
return haze, gt, haze_resize, gt_resize, name
def __len__(self):
return len(self.img_name_list)
class OIHaze_T(data.Dataset):
def __init__(self, root, mode, suffix=None, crop=None, flip=False, resize=1024):
self.root = root
assert mode in ['train', 'test']
if mode == 'train':
self.img_name_list = make_dataset_oihaze_train_triple(root, suffix)
else:
self.img_name_list = make_dataset_oihaze_test(root)
self.mode = mode
self.crop = crop
self.flip = flip
self.resize = transforms.Resize(resize)
def __getitem__(self, index):
if self.mode == 'train':
haze_path, gt_path, predict_path = self.img_name_list[index]
name = os.path.splitext(os.path.split(haze_path)[1])[0]
haze = Image.open(haze_path).convert('RGB')
gt = Image.open(gt_path).convert('RGB')
predict = Image.open(predict_path).convert('RGB')
if self.crop:
haze, gt, predict = random_crop(self.crop, haze, gt, predict)
if self.flip and random.random() < 0.5:
haze = haze.transpose(Image.FLIP_LEFT_RIGHT)
gt = gt.transpose(Image.FLIP_LEFT_RIGHT)
predict = predict.transpose(Image.FLIP_LEFT_RIGHT)
haze, gt, predict = to_tensor(haze), to_tensor(gt), to_tensor(predict)
return haze, gt, predict, name
else:
haze_path, gt_path = self.img_name_list[index]
name = os.path.splitext(os.path.split(haze_path)[1])[0]
haze = Image.open(haze_path).convert('RGB')
gt = Image.open(gt_path).convert('RGB')
haze_resize = self.resize(haze)
haze, gt, haze_resize = to_tensor(haze), to_tensor(gt), to_tensor(haze_resize)
return haze, gt, haze_resize, name
def __len__(self):
return len(self.img_name_list)
class OIHaze2(data.Dataset):
def __init__(self, root, mode, suffix=None, flip=False, crop=None, scale=None, rotate=None):
assert mode in ['train', 'test']
self.root = root
self.mode = mode
if mode == 'train':
self.img_name_list = make_dataset_oihaze_train(root, suffix)
else:
self.img_name_list = make_dataset_oihaze_test(root)
self.scale = transforms.Resize(scale)
self.flip = flip
self.crop = crop
self.rotate = rotate
def __getitem__(self, index):
haze_path, gt_path = self.img_name_list[index]
name = os.path.splitext(os.path.split(haze_path)[1])[0]
haze = Image.open(haze_path).convert('RGB')
gt = Image.open(gt_path).convert('RGB')
if self.mode == 'test':
haze_lr = self.scale(haze)
haze_lr = to_tensor(haze_lr)
else:
if self.rotate:
rotate_degree = random.random() * 2 * self.rotate - self.rotate
haze = haze.rotate(rotate_degree, Image.BILINEAR)
gt = gt.rotate(rotate_degree, Image.BILINEAR)
if self.crop:
haze, gt = random_crop(self.crop, haze, gt)
if self.flip and random.random() < 0.5:
haze = haze.transpose(Image.FLIP_LEFT_RIGHT)
gt = gt.transpose(Image.FLIP_LEFT_RIGHT)
haze = to_tensor(haze)
gt = to_tensor(gt)
if self.mode == 'test':
return haze, gt, haze_lr, name
else:
return haze, gt, name
def __len__(self):
return len(self.img_name_list)
class OIHaze2_2(data.Dataset):
def __init__(self, root, mode, flip=False, crop=None):
assert mode in ['train', 'test']
self.root = root
self.mode = mode
if mode == 'train':
self.img_name_list = make_dataset_oihaze_train(root)
else:
self.img_name_list = make_dataset_oihaze_test(root)
self.flip = flip
self.crop = crop
def __getitem__(self, index):
haze_path, gt_path = self.img_name_list[index]
name = os.path.splitext(os.path.split(haze_path)[1])[0]
haze = Image.open(haze_path).convert('RGB')
gt = Image.open(gt_path).convert('RGB')
if self.mode == 'test':
haze_lr = haze.resize((1024, 1024), resample=Image.BILINEAR)
haze_lr = to_tensor(haze_lr)
if self.crop:
haze, gt = random_crop(self.crop, haze, gt)
haze = to_tensor(haze)
gt = to_tensor(gt)
if self.mode == 'test':
return haze, gt, haze_lr, name
else:
return haze, gt, name
def __len__(self):
return len(self.img_name_list)
class OIHaze4(data.Dataset):
def __init__(self, root, mode, crop=None):
assert mode in ['train', 'test']
self.root = root
self.mode = mode
if mode == 'train':
self.img_name_list = make_dataset_oihaze_train(root)
else:
self.img_name_list = make_dataset_oihaze_test(root)
self.crop = crop
def __getitem__(self, index):
haze_path, gt_path = self.img_name_list[index]
name = os.path.splitext(os.path.split(haze_path)[1])[0]
haze = Image.open(haze_path).convert('RGB')
gt = Image.open(gt_path).convert('RGB')
if self.mode == 'train':
if self.crop:
haze, gt = random_crop(self.crop, haze, gt)
else:
haze_512 = to_tensor(transforms.Resize(512)(haze))
haze_1024 = to_tensor(transforms.Resize(1024)(haze))
haze_2048 = to_tensor(transforms.Resize(2048)(haze))
haze = to_tensor(haze)
gt = to_tensor(gt)
if self.mode == 'train':
return haze, gt, name
else:
return haze, gt, haze_512, haze_1024, haze_2048, name
def __len__(self):
return len(self.img_name_list)
class OIHaze3(data.Dataset):
def __init__(self, root):
self.root = root
self.img_name_list = make_dataset_oihaze_test(root)
def __getitem__(self, index):
haze_path, gt_path = self.img_name_list[index]
name = os.path.splitext(os.path.split(haze_path)[1])[0]
haze = Image.open(haze_path).convert('RGB')
gt = Image.open(gt_path).convert('RGB')
resize = transforms.Resize(512)
haze_lr = resize(haze)
haze_lr = to_tensor(haze_lr)
haze = to_tensor(haze)
gt = to_tensor(gt)
return haze, gt, haze_lr, name
def __len__(self):
return len(self.img_name_list)
class ImageFolder3(data.Dataset):
def __init__(self, root):
self.root = root
self.imgs = [os.path.join(root, img_name) for img_name in os.listdir(root)]
def __getitem__(self, index):
haze_path = self.imgs[index]
name = os.path.splitext(os.path.split(haze_path)[1])[0]
haze = Image.open(haze_path).convert('RGB')
haze = to_tensor(haze)
return haze, name
def __len__(self):
return len(self.imgs)