forked from echonax07/MMSeaIce
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathloaders.py
878 lines (707 loc) · 39.4 KB
/
loaders.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
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""Pytorch Dataset class for training. Function used in train.py."""
# -- File info -- #
__author__ = 'Muhammed Patel'
__contributor__ = 'Xinwwei chen, Fernando Pena Cantu,Javier Turnes, Eddie Park'
__copyright__ = ['university of waterloo']
__contact__ = ['[email protected]', '[email protected]']
__version__ = '1.0.0'
__date__ = '2024-04-05'
# -- Built-in modules -- #
import os
import datetime
from dateutil import relativedelta
import re
from tqdm import tqdm
# -- Third-party modules -- #
import numpy as np # type: ignore
import torch # type: ignore
import xarray as xr # type: ignore
from torch.utils.data import Dataset # type: ignore
import torchvision.transforms.functional as TF # type: ignore
from scipy.ndimage import maximum_filter, minimum_filter # type: ignore
# -- Proprietary modules -- #
from functions import rand_bbox, fill_gaps
amsr2_dims = ('2km_grid_lines', '2km_grid_samples')
sar_size_dims = ("sar_lines", "sar_samples")
def load_netcdf(file):
try:
scene = xr.open_dataset(file, engine='h5netcdf')
except Exception as inst:
print(file) # the exception type
print(type(inst)) # the exception type
print(inst.args) # arguments stored in .args
print(inst)
return None
return scene
def load_npz(file):
try:
with np.load(file) as ds:
scene_dict = {}
for name in ds.files:
if name.startswith('btemp_'):
scene_dict[name] = (amsr2_dims, ds[name])
else:
scene_dict[name] = (sar_size_dims, ds[name])
scene = xr.Dataset(scene_dict)
except Exception as inst:
print(file) # the exception type
print(type(inst)) # the exception type
print(inst.args) # arguments stored in .args
print(inst)
return None
return scene
class AI4ArcticChallengeDataset(Dataset):
"""Pytorch dataset for loading batches of patches of scenes from the ASID
V2 data set."""
def __init__(self, options, files, do_transform=False):
self.options = options
self.files = files
self.do_transform = do_transform
# If Downscaling, down sample data and put in on memory
#if (self.options['down_sample_scale'] == 1):
# self.downsample = False
#else:
# self.downsample = True
self.downsample = True
if self.downsample:
self.scenes = []
self.amsrs = []
self.aux = []
for file in tqdm(self.files):
filepath = os.path.join(self.options['path_to_train_data'], file)
if filepath.endswith('.npz'):
scene = load_npz(filepath)
else:
scene = load_netcdf(filepath)
if scene is None:
continue
try:
temp_scene = scene[self.options['full_variables']].to_array().to_numpy()
except Exception as inst:
print(file) # the exception type
print(type(inst)) # the exception type
print(inst.args) # arguments stored in .args
print(inst)
continue
if 'mask_classes' in self.options:
for i, key in enumerate(self.options['full_variables']):
if key in self.options['mask_classes']:
temp_scene[i] = np.where(temp_scene[i] == self.options['mask_classes'][key], 255, temp_scene[i])
if 'join_classes' in self.options:
for i, chart in enumerate(self.options['join_classes']):
temp_scene_i = np.zeros(temp_scene[i].shape, np.uint8) + np.uint8(255)
for new_class in self.options['join_classes'][chart]:
temp_scene_i[np.isin(temp_scene[i], self.options['join_classes'][chart][new_class])] = new_class
temp_scene[i] = temp_scene_i
if 'change_SIR_footprint' in self.options:
mask = temp_scene[0] == 255
if self.options['change_SIR_footprint'] < 0:
mask = maximum_filter(mask, size=-self.options['change_SIR_footprint'])
temp_scene[0, mask] = 255
else:
temp_scene[0] = fill_gaps(temp_scene[0], mask, self.options['change_SIR_footprint'])
temp_scene = torch.from_numpy(np.expand_dims(temp_scene, 0))
temp_scene = torch.nn.functional.interpolate(temp_scene,
size=(temp_scene.size(2)//self.options['down_sample_scale'],
temp_scene.size(3)//self.options['down_sample_scale']),
mode=self.options['loader_downsampling'])
scene_size_before_padding = temp_scene.shape
if temp_scene.size(2) < self.options['patch_size']:
height_pad = self.options['patch_size'] - temp_scene.size(2) + 1
else:
height_pad = 0
if temp_scene.size(3) < self.options['patch_size']:
width_pad = self.options['patch_size'] - temp_scene.size(3) + 1
else:
width_pad = 0
if height_pad > 0 or width_pad > 0:
temp_scene_y = torch.nn.functional.pad(
temp_scene[:, :len(self.options['charts'])], (0, width_pad, 0, height_pad), mode='constant', value=255)
temp_scene_x = torch.nn.functional.pad(
temp_scene[:, len(self.options['charts']):], (0, width_pad, 0, height_pad), mode='constant', value=0)
temp_scene = torch.cat((temp_scene_y, temp_scene_x), dim=1)
if len(self.options['amsrenv_variables']) > 0:
temp_amsr = np.array(scene[self.options['amsrenv_variables']].to_array())
self.amsrs.append(temp_amsr)
if len(self.options['auxiliary_variables']) > 0:
temp_aux = []
if 'aux_time' in self.options['auxiliary_variables']:
# Get Scene time
scene_id = scene.attrs['scene_id']
# Convert Scene time to number data
norm_time = get_norm_month(scene_id)
time_array = torch.from_numpy(
np.full((scene_size_before_padding[2], scene_size_before_padding[3]), norm_time)).unsqueeze(0).unsqueeze(0)
# time_array = torch.nn.functional.interpolate(time_array.unsqueeze(0).unsqueeze(0),
# size=(scene_size_before_padding.size(2),
# scene_size_before_padding.size(3)),
# mode=self.options['loader_upsampling'])
if height_pad > 0 or width_pad > 0:
time_array = torch.nn.functional.pad(
time_array, (0, width_pad, 0, height_pad), mode='constant', value=0)
temp_aux.append(time_array)
if 'aux_lat' in self.options['auxiliary_variables']:
# Get Latitude
lat_array = scene['sar_grid2d_latitude'].values
lat_array = (lat_array - self.options['latitude']['mean'])/self.options['latitude']['std']
# Interpolate to size of original scene
inter_lat_array = torch.nn.functional.interpolate(input=torch.from_numpy(lat_array).view((1, 1, lat_array.shape[0], lat_array.shape[1])),
size=(scene_size_before_padding[2],
scene_size_before_padding[3]),
mode=self.options['loader_upsampling'])
if height_pad > 0 or width_pad > 0:
inter_lat_array = torch.nn.functional.pad(
inter_lat_array, (0, width_pad, 0, height_pad), mode='constant', value=0)
temp_aux.append(inter_lat_array)
if 'aux_long' in self.options['auxiliary_variables']:
# Get Longuitude
long_array = scene['sar_grid2d_longitude'].values
long_array = (long_array - self.options['longitude']['mean'])/self.options['longitude']['std']
# Interpolate to size of original scene
inter_long_array = torch.nn.functional.interpolate(input=torch.from_numpy(long_array).view((1, 1, lat_array.shape[0], lat_array.shape[1])),
size=(scene_size_before_padding[2],
scene_size_before_padding[3]),
mode=self.options['loader_upsampling'])
if height_pad > 0 or width_pad > 0:
inter_long_array = torch.nn.functional.pad(
inter_long_array, (0, width_pad, 0, height_pad), mode='constant', value=0)
temp_aux.append(inter_long_array)
self.aux.append(torch.cat(temp_aux, 1))
temp_scene = torch.squeeze(temp_scene)
self.scenes.append(temp_scene)
# Channel numbers in patches, includes reference channel.
self.patch_c = len(
self.options['train_variables']) + len(self.options['charts'])
def __len__(self):
"""
Provide number of iterations per epoch. Function required by Pytorch
dataset.
Returns
-------
Number of iterations per epoch.
"""
return self.options['epoch_len']
def random_crop(self, scene):
"""
Perform random cropping in scene.
Parameters
----------
scene :
Xarray dataset; a scene from ASID3 ready-to-train challenge
dataset.
Returns
-------
x_patch :
torch array with shape (len(train_variables),
patch_height, patch_width). None if empty patch.
y_patch :
torch array with shape (len(charts),
patch_height, patch_width). None if empty patch.
"""
patch = np.zeros((len(self.options['full_variables']) +
len(self.options['amsrenv_variables']) +
len(self.options['auxiliary_variables']),
self.options['patch_size'],
self.options['patch_size']))
# Get random index to crop from.
row_rand = np.random.randint(
low=0, high=scene['SIC'].values.shape[0]
- self.options['patch_size'])
col_rand = np.random.randint(
low=0, high=scene['SIC'].values.shape[1]
- self.options['patch_size'])
# Equivalent in amsr and env variable grid.
amsrenv_row = row_rand / self.options['amsrenv_delta']
# Used in determining the location of the crop in between pixels.
amsrenv_row_dec = int(amsrenv_row - int(amsrenv_row))
amsrenv_row_index_crop = amsrenv_row_dec * self.options['amsrenv_delta'] * amsrenv_row_dec
amsrenv_col = col_rand / self.options['amsrenv_delta']
amsrenv_col_dec = int(amsrenv_col - int(amsrenv_col))
amsrenv_col_index_crop = amsrenv_col_dec * self.options['amsrenv_delta'] * amsrenv_col_dec
# - Discard patches with too many meaningless pixels (optional).
if np.sum(scene['SIC'].values[row_rand: row_rand + self.options['patch_size'],
col_rand: col_rand + self.options['patch_size']]
!= self.options['class_fill_values']['SIC']) > 1:
# Crop full resolution variables.
patch[0:len(self.options['full_variables']), :, :] = \
scene[self.options['full_variables']].isel(
sar_lines=range(row_rand, row_rand +
self.options['patch_size']),
sar_samples=range(col_rand, col_rand
+ self.options['patch_size'])).to_array().values
if len(self.options['amsrenv_variables']) > 0:
# Crop and upsample low resolution variables.
patch[len(self.options['full_variables']):len(self.options['full_variables'])+len(self.options['amsrenv_variables']):, :, :] = torch.nn.functional.interpolate(
input=torch.from_numpy(scene[self.options['amsrenv_variables']].to_array().values[
:,
int(amsrenv_row): int(amsrenv_row + np.ceil(self.options['amsrenv_patch'])),
int(amsrenv_col): int(amsrenv_col + np.ceil(self.options['amsrenv_patch']))]
).unsqueeze(0),
size=self.options['amsrenv_upsample_shape'],
mode=self.options['loader_upsampling']).squeeze(0)[
:,
int(np.around(amsrenv_row_index_crop)): int(np.around
(amsrenv_row_index_crop
+ self.options['patch_size'])),
int(np.around(amsrenv_col_index_crop)): int(np.around
(amsrenv_col_index_crop
+ self.options['patch_size']))].numpy()
# Only add auxiliary_variables if they are called
if len(self.options['auxiliary_variables']) > 0:
aux_feat_list = []
if 'aux_time' in self.options['auxiliary_variables']:
# Get Scene time
scene_id = scene.attrs['scene_id']
# Convert Scene time to number data
norm_time = get_norm_month(scene_id)
#
time_array = np.full((self.options['patch_size'],
self.options['patch_size']), norm_time)
aux_feat_list.append(time_array)
if 'aux_lat' in self.options['auxiliary_variables']:
# Get Latitude
lat_array = scene['sar_grid2d_latitude'].values
lat_array = (lat_array - self.options['latitude']['mean'])/self.options['latitude']['std']
# Interpolate to size of original scene
inter_lat_array = torch.nn.functional.interpolate(input=torch.from_numpy(lat_array).view((1, 1, lat_array.shape[0], lat_array.shape[1])), size=scene['nersc_sar_primary'].values.shape,
mode=self.options['loader_upsampling']).numpy()
# Crop to correct patch size
crop_inter_lat_array = inter_lat_array[0, 0, row_rand: row_rand + self.options['patch_size'],
col_rand: col_rand + self.options['patch_size']]
# Append to array
aux_feat_list.append(crop_inter_lat_array)
if 'aux_long' in self.options['auxiliary_variables']:
# Get Longuitude
long_array = scene['sar_grid2d_longitude'].values
long_array = (long_array - self.options['longitude']['mean'])/self.options['longitude']['std']
# Interpolate to size of original scene
inter_long_array = torch.nn.functional.interpolate(input=torch.from_numpy(long_array).view((1, 1, lat_array.shape[0], lat_array.shape[1])), size=scene['nersc_sar_primary'].values.shape,
mode=self.options['loader_upsampling']).numpy()
# Crop to correct patch size
crop_inter_long_array = inter_long_array[0, 0, row_rand: row_rand + self.options['patch_size'],
col_rand: col_rand + self.options['patch_size']]
# Append to array
aux_feat_list.append(crop_inter_long_array)
aux_np_array = np.stack(aux_feat_list, axis=0)
patch[len(self.options['full_variables']) + len(self.options['amsrenv_variables']):, :, :] = aux_np_array
# Separate in to x (train variables) and y (targets) and downscale if needed
x_patch = torch.from_numpy(
patch[len(self.options['charts']):, :]).type(torch.float).unsqueeze(0)
# The following code was commented because down_scale no longer happens here
# if (self.options['down_sample_scale'] != 1):
# x_patch = torch.nn.functional.interpolate(
# x, scale_factor=1/self.options['down_sample_scale'], mode=self.options['loader_downsampling'])
y_patch = torch.from_numpy(patch[:len(self.options['charts']), :, :]).unsqueeze(0)
# The following code was commented because down_scale no longer happens here
# if (self.options['down_sample_scale'] != 1):
# y_patch = torch.nn.functional.interpolate(
# y, scale_factor=1/self.options['down_sample_scale'], mode='nearest')
# In case patch does not contain any valid pixels - return None.
else:
x_patch = None
y_patch = None
return x_patch, y_patch
def random_crop_downsample(self, idx):
"""
Perform random cropping in scene.
Parameters
----------
idx :
Index from self.files to parse
Returns
-------
patch :
Numpy array with shape (len(train_variables),
patch_height, patch_width). None if empty patch.
"""
patch = np.zeros((len(self.options['full_variables']) +
len(self.options['amsrenv_variables']) +
len(self.options['auxiliary_variables']),
self.options['patch_size'],
self.options['patch_size']))
# Get random index to crop from.
row_rand = np.random.randint(
low=0, high=self.scenes[idx].size(1)
- self.options['patch_size'])
col_rand = np.random.randint(
low=0, high=self.scenes[idx].size(2)
- self.options['patch_size'])
# Equivalent in amsr and env variable grid.
amsrenv_row = row_rand / self.options['amsrenv_delta']
# Used in determining the location of the crop in between pixels.
amsrenv_row_dec = int(amsrenv_row - int(amsrenv_row))
amsrenv_row_index_crop = amsrenv_row_dec * self.options['amsrenv_delta'] * amsrenv_row_dec
amsrenv_col = col_rand / self.options['amsrenv_delta']
amsrenv_col_dec = int(amsrenv_col - int(amsrenv_col))
amsrenv_col_index_crop = amsrenv_col_dec * self.options['amsrenv_delta'] * amsrenv_col_dec
# - Discard patches with too many meaningless pixels (optional).
if 'mask_for_cropping' in self.options:
mask_var_index = self.options['mask_for_cropping']['index']
mask_var_chart = self.options['mask_for_cropping']['chart']
else:
mask_var_index = 0
mask_var_chart = 'SIC'
if np.sum(self.scenes[idx][mask_var_index, row_rand: row_rand + self.options['patch_size'],
col_rand: col_rand + self.options['patch_size']].numpy()
!= self.options['class_fill_values'][mask_var_chart]) > 100:
# Crop full resolution variables.
patch[0:len(self.options['full_variables']), :, :] = \
self.scenes[idx][:, row_rand:row_rand + int(self.options['patch_size']),
col_rand:col_rand + int(self.options['patch_size'])].numpy()
if len(self.options['amsrenv_variables']) > 0:
# Crop and upsample low resolution variables.
amsrenv = torch.from_numpy(self.amsrs[idx][:,
int(amsrenv_row): int(amsrenv_row + np.ceil(self.options['amsrenv_patch'])),
int(amsrenv_col): int(amsrenv_col + np.ceil(self.options['amsrenv_patch']))]
).unsqueeze(0)
# Add padding in case the patch size return is smaller than the expected one.
if amsrenv.size(2) < self.options['amsrenv_patch']:
height_pad = int(np.ceil(self.options['amsrenv_patch'])) - amsrenv.size(2)
else:
height_pad = 0
if amsrenv.size(3) < self.options['amsrenv_patch']:
width_pad = int(np.ceil(self.options['amsrenv_patch'])) - amsrenv.size(3)
else:
width_pad = 0
if height_pad > 0 or width_pad > 0:
amsrenv = torch.nn.functional.pad(amsrenv, (0, width_pad, 0, height_pad), mode='constant', value=0)
# TODO The square bracket part is redundant []. for Example if size=2560 then doing [0:2560] after interpolate is redundant
amsrenv = torch.nn.functional.interpolate(
input=amsrenv,
size=self.options['amsrenv_upsample_shape'],
mode=self.options['loader_upsampling']).squeeze(0)[
:,
int(np.around(amsrenv_row_index_crop)): int(np.around
(amsrenv_row_index_crop
+ self.options['patch_size'])),
int(np.around(amsrenv_col_index_crop)): int(np.around
(amsrenv_col_index_crop
+ self.options['patch_size']))]
patch[len(self.options['full_variables']):len(self.options['full_variables']) +
len(self.options['amsrenv_variables']):, :, :] = amsrenv.numpy()
# Only add auxiliary_variables if they are called
# No need to do the patch
if len(self.options['auxiliary_variables']) > 0:
patch[len(self.options['full_variables']) + len(self.options['amsrenv_variables']):, :, :] = self.aux[idx][0, :, row_rand: row_rand +
self.options['patch_size'], col_rand: col_rand + self.options['patch_size']]
x_patch = torch.from_numpy(
patch[len(self.options['charts']):, :]).type(torch.float).unsqueeze(0)
y_patch = torch.from_numpy(patch[:len(self.options['charts']), :, :]).unsqueeze(0)
# In case patch does not contain any valid pixels - return None.
else:
x_patch = None
y_patch = None
return x_patch, y_patch
def prep_dataset(self, x_patches, y_patches):
"""
Convert patches from 4D numpy array to 4D torch tensor.
Parameters
----------
x_patches : ndarray
Patches sampled from ASID3 ready-to-train challenge dataset scenes [PATCH, CHANNEL, H, W] containing only the trainable variables.
y_patches : ndarray
Patches sampled from ASID3 ready-to-train challenge dataset scenes [PATCH, CHANNEL, H, W] contrainng only the targets.
Returns
-------
x :
4D torch tensor; ready training data.
y : Dict
Dictionary with 3D torch tensors for each chart; reference data for training data x.
"""
# Convert training data to tensor float.
x = x_patches.type(torch.float)
# Store charts in y dictionary.
y = {}
for idx, chart in enumerate(self.options['charts']):
y[chart] = y_patches[:, idx].type(torch.long)
return x, y
def transform(self, x_patch, y_patch):
data_aug_options = self.options['data_augmentations']
if torch.rand(1) < data_aug_options['Random_h_flip']:
x_patch = TF.hflip(x_patch)
y_patch = TF.hflip(y_patch)
if torch.rand(1) < data_aug_options['Random_v_flip']:
x_patch = TF.vflip(x_patch)
y_patch = TF.vflip(y_patch)
assert (data_aug_options['Random_rotation'] <= 180)
if data_aug_options['Random_rotation'] != 0 and \
torch.rand(1) < data_aug_options['Random_rotation_prob']:
random_degree = np.random.randint(-data_aug_options['Random_rotation'],
data_aug_options['Random_rotation']
)
else:
random_degree = 0
scale_diff = data_aug_options['Random_scale'][1] - \
data_aug_options['Random_scale'][0]
assert (scale_diff >= 0)
if scale_diff != 0 and torch.rand(1) < data_aug_options['Random_scale_prob']:
random_scale = np.random.rand()*(data_aug_options['Random_scale'][1] -
data_aug_options['Random_scale'][0]) +\
data_aug_options['Random_scale'][0]
else:
random_scale = data_aug_options['Random_scale'][1]
x_patch = TF.affine(x_patch, angle=random_degree, translate=(0, 0),
shear=0, scale=random_scale, fill=0)
y_patch = TF.affine(y_patch, angle=random_degree, translate=(0, 0),
shear=0, scale=random_scale, fill=255)
return x_patch, y_patch
def __getitem__(self, idx):
"""
Get batch. Function required by Pytorch dataset.
Returns
-------
x :
4D torch tensor; ready training data.
y : Dict
Dictionary with 3D torch tensors for each chart; reference data for training data x.
"""
# Placeholder to fill with data.
x_patches = torch.zeros((self.options['batch_size'], len(self.options['train_variables']),
self.options['patch_size'], self.options['patch_size']))
y_patches = torch.zeros((self.options['batch_size'], len(self.options['charts']),
self.options['patch_size'], self.options['patch_size']))
sample_n = 0
# Continue until batch is full.
while sample_n < self.options['batch_size']:
# - Open memory location of scene. Uses 'Lazy Loading'.
scene_id = np.random.randint(
low=0, high=len(self.files), size=1).item()
# - Extract patches
try:
if self.downsample:
x_patch, y_patch = self.random_crop_downsample(scene_id)
else:
scene = xr.open_dataset(os.path.join(
self.options['path_to_train_data'], self.files[scene_id]), engine='h5netcdf')
x_patch, y_patch = self.random_crop(scene)
except Exception as e:
if self.downsample:
print(f"Cropping in {self.files[scene_id]} failed.")
print(f"Scene size: {self.scenes[scene_id][0].shape} for crop shape: \
({self.options['patch_size']}, {self.options['patch_size']})")
print('Skipping scene.')
continue
else:
print(f"Cropping in {self.files[scene_id]} failed.")
print(f"Scene size: {scene['SIC'].values.shape} for crop shape: \
({self.options['patch_size']}, {self.options['patch_size']})")
print('Skipping scene.')
continue
if x_patch is not None:
if self.do_transform:
try:
x_patch, y_patch = self.transform(x_patch, y_patch)
except:
print('Error with self.transform.')
continue
# -- Stack the scene patches in patches
x_patches[sample_n, :, :, :] = x_patch
y_patches[sample_n, :, :, :] = y_patch
sample_n += 1 # Update the index.
if self.do_transform and torch.rand(1) < self.options['data_augmentations']['Cutmix_prob']:
lam = np.random.beta(self.options['data_augmentations']['Cutmix_beta'],
self.options['data_augmentations']['Cutmix_beta'])
rand_index = torch.randperm(x_patches.size(0))
bbx1, bby1, bbx2, bby2 = rand_bbox(x_patches.size(), lam)
x_patches[:, :, bbx1:bbx2, bby1:bby2] = x_patches[rand_index, :, bbx1:bbx2, bby1:bby2]
y_patches[:, :, bbx1:bbx2, bby1:bby2] = y_patches[rand_index, :, bbx1:bbx2, bby1:bby2]
# Prepare training arrays
x, y = self.prep_dataset(x_patches, y_patches)
return x, y
class AI4ArcticChallengeTestDataset(Dataset):
"""Pytorch dataset for loading full scenes from the ASID ready-to-train challenge dataset for inference."""
def __init__(self, options, files, mode='test'):
self.options = options
self.files = files
# if mode not in ["train_val", "test_val", "test"]:
if mode not in ["train", "test", "test_no_gt"]:
raise ValueError("String variable must be one of 'train', 'test', or 'test_no_gt'")
self.mode = mode
self.data = []
for i, ifile in enumerate(tqdm(self.files)):
x, y, cfv_masks, tfv_mask, name, original_size = self._getitem_old(i)
if x is None:
continue
self.data.append([x, y, cfv_masks, tfv_mask, name, original_size])
def __getitem__(self, idx):
return self.data[idx]
def __len__(self):
"""
Provide the number of iterations. Function required by Pytorch dataset.
Returns
-------
Number of scenes per validation.
"""
return len(self.files)
def prep_scene(self, scene):
"""
Upsample low resolution to match charts and SAR resolution. Convert patches
from 4D numpy array to 4D torch tensor.
Parameters
----------
scene :
Returns
-------
x :
4D torch tensor, ready training data.
y :
Dict with 3D torch tensors for each reference chart; reference inference data for x. None if test is true.
"""
x_feat_list = []
sar_var_x = torch.from_numpy(
scene[self.options['sar_variables']].to_array().values).unsqueeze(0)
x_feat_list.append(sar_var_x)
size = scene['nersc_sar_primary'].values.shape
if len(self.options['amsrenv_variables']) > 0:
# from icecream import ic
# print(1, scene['SIC'].values.shape)
# print(2, scene['nersc_sar_primary'].values.shape)
asmr_env__var_x = torch.nn.functional.interpolate(input=torch.from_numpy(
scene[self.options['amsrenv_variables']].to_array().values).unsqueeze(0),
size=size,
mode=self.options['loader_upsampling'])
x_feat_list.append(asmr_env__var_x)
# Only add auxiliary_variables if they are called
if len(self.options['auxiliary_variables']) > 0:
if 'aux_time' in self.options['auxiliary_variables']:
# Get Scene time
scene_id = scene.attrs['scene_id']
# Convert Scene time to number data
norm_time = get_norm_month(scene_id)
#
time_array = torch.from_numpy(
np.full(scene['nersc_sar_primary'].values.shape, norm_time)).view(1, 1, size[0], size[1])
x_feat_list.append(time_array,)
if 'aux_lat' in self.options['auxiliary_variables']:
# Get Latitude
lat_array = scene['sar_grid2d_latitude'].values
lat_array = (lat_array - self.options['latitude']['mean'])/self.options['latitude']['std']
# Interpolate to size of original scene
inter_lat_array = torch.nn.functional.interpolate(input=torch.from_numpy(lat_array).view((1, 1, lat_array.shape[0], lat_array.shape[1])), size=size,
mode=self.options['loader_upsampling'])
# Append to array
x_feat_list.append(inter_lat_array)
if 'aux_long' in self.options['auxiliary_variables']:
# Get Longitude
long_array = scene['sar_grid2d_longitude'].values
long_array = (long_array - self.options['longitude']['mean'])/self.options['longitude']['std']
# Interpolate to size of original scene
inter_long_array = torch.nn.functional.interpolate(input=torch.from_numpy(long_array).view((1, 1, lat_array.shape[0], lat_array.shape[1])), size=size,
mode=self.options['loader_upsampling'])
# Append to array
x_feat_list.append(inter_long_array)
# x_feat_list.append(aux_var_x)
x = torch.cat(x_feat_list, axis=1)
# else:
# x = torch.from_numpy(
# scene[self.options['sar_variables']].to_array().values).unsqueeze(0)
# Downscale if needed
if (self.options['down_sample_scale'] != 1):
x = torch.nn.functional.interpolate(
x, scale_factor=1/self.options['down_sample_scale'], mode=self.options['loader_downsampling'])
# TODO:
if self.mode != 'test_no_gt':
y_charts = torch.from_numpy(scene[self.options['charts']].isel().to_array().values).unsqueeze(0)
y_charts = torch.nn.functional.interpolate(
y_charts, scale_factor=1/self.options['down_sample_scale'], mode='nearest')
y = {}
for idx, chart in enumerate(self.options['charts']):
y[chart] = y_charts[:, idx].squeeze().numpy()
if 'join_classes' in self.options:
for chart in self.options['join_classes']:
y_chart = np.zeros(y[chart].shape, np.uint8) + np.uint8(255)
for new_class in self.options['join_classes'][chart]:
y_chart[np.isin(y[chart], self.options['join_classes'][chart][new_class])] = new_class
y[chart] = y_chart
else:
y = None
return x.float(), y
def _getitem_old(self, idx):
"""
Get scene. Function required by Pytorch dataset.
Returns
-------
x :
4D torch tensor; ready inference data.
y :
Dict with 3D torch tensors for each reference chart; reference inference data for x. None if test is true.
masks :
Dict with 2D torch tensors; mask for each chart for loss calculation. Contain only SAR mask if test is true.
name : str
Name of scene.
"""
if self.mode == 'test' or self.mode == 'test_no_gt':
filepath = os.path.join(self.options['path_to_test_data'], self.files[idx])
elif self.mode == 'train':
filepath = os.path.join(self.options['path_to_train_data'], self.files[idx])
if filepath.endswith('.npz'):
scene = load_npz(filepath)
elif filepath.endswith('.nc'):
scene = load_netcdf(filepath)
if scene is None:
return None, None, None, None, None, None
x, y = self.prep_scene(scene)
name = self.files[idx]
if self.mode != 'test_no_gt':
cfv_masks = {}
for chart in self.options['charts']:
cfv_masks[chart] = (
(y[chart] == self.options['class_fill_values'][chart]) +
(y[chart] >= self.options['n_classes'][chart]))
if 'mask_classes' in self.options and chart in self.options['mask_classes']:
cfv_masks[chart] += np.isin(y[chart], self.options['mask_classes'][chart])
cfv_masks[chart] = cfv_masks[chart].squeeze()
else:
cfv_masks = None
tfv_mask = (x.squeeze()[0, :, :] ==
self.options['train_fill_value']).squeeze()
original_size = scene['nersc_sar_primary'].values.shape
return x, y, cfv_masks, tfv_mask, name, original_size
def get_variable_options(train_options: dict):
"""
Get amsr and env grid options, crop shape and upsampling shape.
Parameters
----------
train_options: dict
Dictionary with training options.
Returns
-------
train_options: dict
Updated with amsrenv options.
Updated with correct true patch size
"""
train_options['amsrenv_delta'] = train_options['amsrenv_pixel_spacing'] / \
(train_options['pixel_spacing']*train_options['down_sample_scale'])
train_options['amsrenv_patch'] = train_options['patch_size'] / \
train_options['amsrenv_delta']
train_options['amsrenv_patch_dec'] = int(
train_options['amsrenv_patch'] - int(train_options['amsrenv_patch']))
train_options['amsrenv_upsample_shape'] = (int(train_options['patch_size'] +
train_options['amsrenv_patch_dec'] *
train_options['amsrenv_delta']),
int(train_options['patch_size'] +
train_options['amsrenv_patch_dec'] *
train_options['amsrenv_delta']))
train_options['sar_variables'] = [variable for variable in train_options['train_variables']
if 'sar' in variable or 'map' in variable]
train_options['full_variables'] = np.hstack((train_options['charts'], train_options['sar_variables']))
train_options['amsrenv_variables'] = [variable for variable in train_options['train_variables']
if 'sar' not in variable and 'map' not in variable and 'aux' not in variable]
train_options['auxiliary_variables'] = [
variable for variable in train_options['train_variables'] if 'aux' in variable]
return train_options
def get_norm_month(file_name):
pattern = re.compile(r'\d{8}T\d{6}')
# Search for the first match in the string
match = re.search(pattern, file_name)
first_date = match.group(0)
# parse the date string into a datetime object
date = datetime.datetime.strptime(first_date, "%Y%m%dT%H%M%S")
# calculate the number of days between January 1st and the given date
delta = relativedelta.relativedelta(date, datetime.datetime(date.year, 1, 1))
# delta = (date - datetime.datetime(date.year, 1, 1)).days
months = delta.months
norm_months = 2*months/11-1
return norm_months