-
Notifications
You must be signed in to change notification settings - Fork 6
/
utils.py
892 lines (740 loc) · 28.9 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
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
"""
Utility functions.
"""
import os
import time
import glob
from pathlib import Path
import atexit
from itertools import repeat
import math
import logging
import random
import pickle
import csv
import multiprocessing
from tqdm import tqdm
import numpy as np
import torch
import torch.distributed as dist
import hydra
from omegaconf import DictConfig
import torch_geometric as pyg
import trimesh
from plyfile import PlyData
from abspy import VertexGroup, CellComplex
def setup_runner(rank, world_size, master_addr, master_port):
"""
Set up runner for distributed parallelization.
"""
# initialize torch.distributed
os.environ['MASTER_ADDR'] = str(master_addr)
os.environ['MASTER_PORT'] = str(master_port)
dist.init_process_group('nccl', rank=rank, world_size=world_size)
def attach_to_log(level=logging.INFO,
filepath=None,
colors=True,
capture_warnings=True):
"""
Attach a stream handler to all loggers.
Parameters
------------
level : enum (int)
Logging level, like logging.INFO
colors : bool
If True try to use colorlog formatter
capture_warnings: bool
If True capture warnings
filepath: None or str
path to save the logfile
Returns
-------
logger: Logger object
Logger attached with a stream handler
"""
# make sure we log warnings from the warnings module
logging.captureWarnings(capture_warnings)
# create a basic formatter
formatter_file = logging.Formatter(
"[%(asctime)s] %(levelname)-7s (%(filename)s:%(lineno)3s) %(message)s",
"%Y-%m-%d %H:%M:%S")
if colors:
try:
from colorlog import ColoredFormatter
formatter_stream = ColoredFormatter(
("%(log_color)s%(levelname)-8s%(reset)s " +
"%(filename)17s:%(lineno)-4s %(blue)4s%(message)s"),
datefmt=None,
reset=True,
log_colors={'DEBUG': 'cyan',
'INFO': 'green',
'WARNING': 'yellow',
'ERROR': 'red',
'CRITICAL': 'red'})
except ImportError:
formatter_stream = formatter_file
else:
formatter_stream = formatter_file
# if no handler was passed use a StreamHandler
logger = logging.getLogger()
logger.setLevel(level)
if not any([isinstance(handler, logging.StreamHandler) for handler in logger.handlers]):
stream_handler = logging.StreamHandler()
stream_handler.setFormatter(formatter_stream)
logger.addHandler(stream_handler)
if filepath and not any([isinstance(handler, logging.FileHandler) for handler in logger.handlers]):
file_handler = logging.FileHandler(filepath)
file_handler.setFormatter(formatter_file)
logger.addHandler(file_handler)
# set nicer numpy print options
np.set_printoptions(precision=5, suppress=True)
return logger
def edge_index_from_dict(graph_dict):
"""
Convert adjacency dict to edge index.
Parameters
----------
graph_dict: dict
Adjacency dict
Returns
-------
as_tensor: torch.Tensor
Edge index
"""
row, col = [], []
for key, value in graph_dict.items():
row += repeat(key, len(value))
col += value
edge_index = torch.tensor([row, col], dtype=torch.long)
return edge_index
def index_to_mask(index, size):
"""
Convert index to binary mask.
Parameters
----------
index: range Object
Index of 1s
size: int
Size of mask
Returns
-------
as_tensor: torch.Tensor
Binary mask
"""
mask = torch.zeros((size,), dtype=torch.bool)
mask[index] = 1
return mask
def freeze_vram(cuda_devices, timeout=500):
"""
Freeze VRAM for a short time at program exit. For debugging.
Parameters
----------
cuda_devices: list of int
Indices of CUDA devices
timeout: int
Timeout seconds
"""
torch.cuda.empty_cache()
devices_info = os.popen(
'"/usr/bin/nvidia-smi" --query-gpu=memory.total,memory.used --format=csv,nounits,noheader') \
.read().strip().split("\n")
for i, device in enumerate(cuda_devices):
total, used = devices_info[int(device)].split(',')
total = int(total)
used = int(used)
max_mem = int(total * 0.90)
block_mem = max_mem - used
if block_mem > 0:
x = torch.FloatTensor(256, 1024, block_mem).to(torch.device(f'cuda:{i}'))
del x
for _ in tqdm(range(timeout), desc='VRAM freezing'):
time.sleep(1)
def init_device(gpu_ids, register_freeze=False):
"""
Init devices.
Parameters
----------
gpu_ids: list of int
GPU indices to use
register_freeze: bool
Register GPU memory freeze if set True
"""
# set multiprocessing sharing strategy
torch.multiprocessing.set_sharing_strategy('file_system')
# does not work for DP after import torch with PyTorch 2.0, but works for DDP nevertheless
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
os.environ["CUDA_VISIBLE_DEVICES"] = str(gpu_ids)[1:-1]
# raise soft limit from 1024 to 4096 for open files to address RuntimeError: received 0 items of ancdata
import resource
rlimit = resource.getrlimit(resource.RLIMIT_NOFILE)
resource.setrlimit(resource.RLIMIT_NOFILE, (4096, rlimit[1]))
if register_freeze:
atexit.register(freeze_vram, gpu_ids)
class Sampler:
"""
Sampler to sample points from a point cloud.
"""
def __init__(self, strategy, length, ratio, resolutions, duplicate, seed=None):
self.strategy = strategy
self.length = length
self.ratio = ratio
self.resolutions = resolutions
self.duplicate = duplicate
# seed once in initialization
self.seed = seed
def sample(self, data):
with torch.no_grad():
if self.seed is not None:
set_seed(self.seed)
if self.strategy is None:
return data
if self.strategy == 'fps':
return self.farthest_sample(data)
elif self.strategy == 'random':
return self.random_sample(data)
elif self.strategy == 'grid':
return self.grid_sample(data)
else:
raise ValueError(f'Unexpected sampling strategy={self.strategy}.')
def random_sample(self, data):
"""
Random uniform sampling.
"""
# https://pytorch-geometric.readthedocs.io/en/latest/_modules/torch_geometric/transforms/fixed_points.html#FixedPoints
if not self.duplicate:
choice = torch.randperm(data.num_points)[:self.length]
else:
choice = torch.cat(
[torch.randperm(data.num_points) for _ in range(math.ceil(self.length / data.num_points))],
dim=0)[:self.length]
data[f'batch_points_{self.length}'] = data.batch_points[choice]
data[f'points_{self.length}'] = data.points[choice]
return data
def grid_sample(self, data):
"""
Sampling points into fixed-sized voxels.
Each cluster returned is the cluster barycenter.
"""
# https://pytorch-geometric.readthedocs.io/en/latest/_modules/torch_geometric/transforms/grid_sampling.html#GridSampling
for size in self.resolutions:
c = pyg.nn.voxel_grid(data.points, size, data.batch_points, None, None)
_, perm = pyg.nn.pool.consecutive.consecutive_cluster(c)
data[f'batch_points_{size}'] = data.batch_points[perm]
data[f'points_{size}'] = data.points[perm]
return data
def farthest_sample(self, data):
"""
Farthest sampling which iteratively samples the most distant point with regard to the rest points. Inplace.
"""
# https://pytorch-geometric.readthedocs.io/en/latest/modules/nn.html#torch_geometric.nn.pool.fps
perm = pyg.nn.pool.fps(data.points, data.batch_points, ratio=self.ratio)
data[f'batch_points_fps'] = data.batch_points[perm]
data[f'points_fps'] = data.points[perm]
return data
def reverse_translation_and_scale(mesh):
"""
Translation and scale for reverse normalisation of mesh.
"""
bounds = mesh.extents
if bounds.min() == 0.0:
return
# translate to origin
translation = - (mesh.bounds[0] + mesh.bounds[1]) * 0.5
translation = trimesh.transformations.translation_matrix(direction=-translation)
# scale to unit cube
scale = bounds.max()
scale_trafo = trimesh.transformations.scale_matrix(factor=scale)
return translation, scale_trafo
def normalise_mesh(args):
"""
Normalize mesh (or point cloud).
First translation, then scaling, if any.
Parameters
----------
args[0]: input_path: str or Path
Path to input mesh (results from reconstruction, normalised)
args[1]: reference_path: str or Path
Path to reference mesh (used to determine the transformation)
args[2]: output_path: str or Path
Path to output mesh (final reconstruction, with reversed normalisation)
args[3]: force: str
Force loading type ('mesh' or 'scene')
args[4]: offset: list
Coordinate offset (x, y, z)
args[5]: scaling: bool
Switch on scaling if set True
args[6]: translation: bool
Switch on translation if set True
"""
input_path, reference_path, output_path, force, offset, is_scaling, is_translation = args
reference_mesh = trimesh.load(reference_path)
translation, scale_trafo = reverse_translation_and_scale(reference_mesh)
if offset is not None:
translation[0][-1] = translation[0][-1] + offset[0]
translation[1][-1] = translation[1][-1] + offset[1]
translation[2][-1] = translation[2][-1] + offset[2]
# trimesh built-in transform would result in an issue of missing triangles
with open(input_path, 'r') as fin:
lines = fin.readlines()
lines_ = []
for line in lines:
if line.startswith('v'):
vertices = np.array(line.split()[1:], dtype=float)
if is_translation is True:
vertices = vertices - translation[:3, 3]
if is_scaling is True:
vertices = vertices / scale_trafo[0][0]
line_ = f'v {vertices[0]} {vertices[1]} {vertices[2]}\n'
else:
line_ = line
lines_.append(line_)
with open(output_path, 'w') as fout:
fout.writelines(lines_)
def reverse_normalise_mesh(args):
"""
Reverse normalisation for reconstructed mesh.
First scaling, then translation, if any.
Parameters
----------
args[0]: input_path: str or Path
Path to input mesh (results from reconstruction, normalised)
args[1]: reference_path: str or Path
Path to reference mesh (used to determine the transformation)
args[2]: output_path: str or Path
Path to output mesh (final reconstruction, with reversed normalisation)
args[3]: force: str
Force loading type ('mesh' or 'scene')
args[4]: offset: list
Coordinate offset (x, y, z)
args[5]: scaling: bool
Switch on scaling if set True
args[6]: translation: bool
Switch on translation if set True
"""
input_path, reference_path, output_path, force, offset, is_scaling, is_translation = args
reference_mesh = trimesh.load(reference_path)
translation, scale_trafo = reverse_translation_and_scale(reference_mesh)
if offset is not None:
translation[0][-1] = translation[0][-1] + offset[0]
translation[1][-1] = translation[1][-1] + offset[1]
translation[2][-1] = translation[2][-1] + offset[2]
# trimesh built-in transform would result in an issue of missing triangles
with open(input_path, 'r') as fin:
lines = fin.readlines()
lines_ = []
for line in lines:
if line.startswith('v'):
vertices = np.array(line.split()[1:], dtype=float)
if is_scaling is True:
vertices = vertices * scale_trafo[0][0]
if is_translation is True:
vertices = vertices + translation[:3, 3]
line_ = f'v {vertices[0]} {vertices[1]} {vertices[2]}\n'
else:
line_ = line
lines_.append(line_)
with open(output_path, 'w') as fout:
fout.writelines(lines_)
def reverse_normalise_cloud(args):
"""
Reverse normalisation for normalised point cloud.
Parameters
----------
args[0]: input_path: str or Path
Path to input point cloud (normalised)
args[1]: reference_path: str or Path
Path to reference mesh (used to determine the transformation)
args[2]: output_path: str or Path
Path to output point cloud (with reversed normalisation)
args[3]: force: str
Force loading type ('mesh' or 'scene')
args[4]: offset: list
Coordinate offset (x, y, z)
"""
input_path, reference_path, output_path, force, offset = args
plydata = PlyData.read(input_path)['vertex']
points = np.array([plydata['x'], plydata['y'], plydata['z']]).T
cloud = trimesh.PointCloud(points)
reference_mesh = trimesh.load(reference_path)
translation, scale_trafo = reverse_translation_and_scale(reference_mesh)
if offset is not None:
translation[0][-1] = translation[0][-1] + offset[0]
translation[1][-1] = translation[1][-1] + offset[1]
translation[2][-1] = translation[2][-1] + offset[2]
cloud.apply_transform(scale_trafo)
cloud.apply_transform(translation)
cloud.export(str(output_path))
def coerce(data):
"""
Coercion for legacy data.
"""
data.points = torch.as_tensor(data.points, dtype=torch.float)
data.queries = torch.as_tensor(data.queries, dtype=torch.float)
if not hasattr(data, 'num_points'):
data.num_points = len(data.points)
if not hasattr(data, 'batch_points'):
data.batch_points = torch.zeros(len(data.points), dtype=torch.long)
return data
def set_seed(seed: int) -> None:
"""
Set singular seed to fix randomness.
May need to be repeatedly invoked (at least for np.random).
"""
np.random.seed(seed)
random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
# When running on the CuDNN backend, two further options must be set
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
# Set a fixed value for the hash seed
os.environ["PYTHONHASHSEED"] = str(seed)
def append_labels(args):
"""
Append occupancy labels to existing cell complex file.
Parameters
----------
args[0]: input_cc: str or Path
Path to input CC file
args[1]: input_manifold: str or Path
Path to input manifold mesh file
args[2]: output_cc: str or Path
Path to output CC file
"""
with open(args[0], 'rb') as handle:
cell_complex = pickle.load(handle)
cells_in_mesh = cell_complex.cells_in_mesh(args[1])
# one-hot encoding
labels = [0] * cell_complex.num_cells
for i in cells_in_mesh:
labels[i] = 1
cell_complex.labels = labels
cell_complex.save(args[2])
@hydra.main(config_path='./conf', config_name='config', version_base='1.2')
def multi_append_labels(cfg: DictConfig):
# initialize logging
logger = logging.getLogger('Labels')
filenames = glob.glob('data/munich_16star/raw/05_complexes/*.cc')
args = []
for filename_input in filenames:
stem = Path(filename_input).stem
filename_output = Path(filename_input).with_suffix('.cc.new')
filename_manifold = Path(filename_input).parent.parent / '03_meshes_manifold' / (stem + '.obj')
if not filename_output.exists():
args.append((filename_input, filename_manifold, filename_output))
logger.info('Start complex labeling')
with multiprocessing.Pool(processes=cfg.num_workers if cfg.num_workers else multiprocessing.cpu_count()) as pool:
# call with multiprocessing
for _ in tqdm(pool.imap_unordered(append_labels, args), desc='Appending labels', total=len(args)):
pass
# exit with a message
logger.info('Done complex labeling')
def append_samples(args):
"""
Append multi-grid-size samples to existing data file.
Parameters
----------
args[0]: input_path: str or Path
Path to input torch file
args[1]: output_path: str or Path
Path to output torch file
args[2]: sample_func: func
Function to sample
"""
data = torch.load(args[0])
data = coerce(data)
data = args[2](data)
torch.save(data, args[1])
def append_queries(args):
"""
Append queries to existing data files.
Parameters
----------
args[0]: input_path: str or Path
Path to input torch file
args[1]: complex_dir: str or Path
Dir to complexes
args[2]: output_path: str or Path
Path to output torch file
"""
data = torch.load(args[0])
with open(os.path.join(args[1], data.name + '.cc'), 'rb') as handle:
cell_complex = pickle.load(handle)
queries_random = np.array(cell_complex.cell_representatives(location='random_t', num=16))
queries_boundary = np.array(cell_complex.cell_representatives(location='boundary', num=16))
queries_skeleton = np.array(cell_complex.cell_representatives(location='skeleton', num=16))
data.queries_random = torch.as_tensor(queries_random, dtype=torch.float)
data.queries_boundary = torch.as_tensor(queries_boundary, dtype=torch.float)
data.queries_skeleton = torch.as_tensor(queries_skeleton, dtype=torch.float)
torch.save(data, args[2])
@hydra.main(config_path='./conf', config_name='config', version_base='1.2')
def multi_append_queries(cfg: DictConfig):
# initialize logging
logger = logging.getLogger('Querying')
filenames = glob.glob(f'{cfg.data_dir}/processed/*[0-9].pt')
args = []
for filename_input in filenames:
filename_output = Path(filename_input).with_suffix('.pt.queries_appended')
if not filename_output.exists():
args.append((filename_input, cfg.complex_dir, filename_output))
logger.info('Start polyhedra sampling')
with multiprocessing.Pool(processes=cfg.num_workers if cfg.num_workers else multiprocessing.cpu_count()) as pool:
# call with multiprocessing
for _ in tqdm(pool.imap_unordered(append_queries, args), desc='Appending queries', total=len(args)):
pass
# exit with a message
logger.info('Done polyhedra sampling')
@hydra.main(config_path='./conf', config_name='config', version_base='1.2')
def multi_append_samples(cfg: DictConfig):
# initialize logging
logger = logging.getLogger('Sampling')
filenames = glob.glob(f'{cfg.data_dir}/processed/*[0-9].pt')
sampler = Sampler(strategy=cfg.sample.strategy, length=cfg.sample.length, ratio=cfg.sample.ratio,
resolutions=cfg.sample.resolutions, duplicate=cfg.sample.duplicate, seed=cfg.seed)
args = []
for filename_input in filenames:
filename_output = Path(filename_input).with_suffix('.pt.samples_appended')
if not filename_output.exists():
args.append((filename_input, filename_output, sampler.sample))
logger.info('Start point cloud sampling')
with multiprocessing.Pool(processes=cfg.num_workers if cfg.num_workers else multiprocessing.cpu_count()) as pool:
# call with multiprocessing
for _ in tqdm(pool.imap_unordered(append_samples, args), desc='Appending samples', total=len(args)):
pass
# exit with a message
logger.info('Done point cloud sampling')
def count_facets(mesh_path):
"""
Count the number of facets given a mesh.
"""
mesh = trimesh.load(mesh_path)
faces_extracted = np.concatenate(mesh.facets)
faces_left = np.setdiff1d(np.arange(len(mesh.faces)), faces_extracted)
num_facets = len(mesh.facets) + len(faces_left)
return num_facets
def dict_count_facets(mesh_dir):
"""
Count the number of facets given a directory of meshes.
"""
filenames = glob.glob(f'{mesh_dir}/*.obj')
facet_dict = {}
for filename_input in filenames:
stem = Path(filename_input).stem
num_facets = count_facets(filename_input)
facet_dict[stem] = num_facets
# sorted by facet number
print({k: v for k, v in sorted(facet_dict.items(), key=lambda item: item[1])})
def append_scale_to_csv(input_csv, output_csv):
"""
Append scale into an existing csv file.
Note that scale has been implemented in stats.py.
"""
rows = []
with open(input_csv, 'r', newline='') as input_csvfile:
reader = csv.reader(input_csvfile)
next(reader) # skip header
for r in reader:
filename_input = r[1]
row = r
if not filename_input.endswith('.obj'):
filename_input = filename_input + '.obj'
mesh = trimesh.load(filename_input)
extents = mesh.extents
scale = extents.max()
row.append(scale)
rows.append(row)
with open(output_csv, 'w') as output_csvfile:
writer = csv.writer(output_csvfile, lineterminator='\n')
writer.writerows(rows)
@hydra.main(config_path='./conf', config_name='config', version_base='1.2')
def calculate(cfg):
"""
Calculate number of parameters.
"""
from network import PolyGNN
# initialize model
model = PolyGNN(cfg)
# calculate params
total_params = sum(
param.numel() for param in model.parameters()
)
trainable_params = sum(
p.numel() for p in model.parameters() if p.requires_grad
)
print(f'total_params: {total_params}')
print(f'trainable_params: {trainable_params}')
def merge_vg2cc(args):
"""
Merge vg from RANSAC and vg from City3D, then generate complexes.
"""
vg_ransac_path, vg_city3d_path, cc_output_path = args
epsilon = 0.0001
vertex_group_ransac = VertexGroup(filepath=vg_ransac_path, refit=True, global_group=False, quiet=True)
vertex_group_city3d = VertexGroup(filepath=vg_city3d_path, refit=False, global_group=True, quiet=True)
additional_planes = [p for p in vertex_group_city3d.planes if -epsilon < p[2] < epsilon or
(-epsilon < p[0] < epsilon and -epsilon < p[1] < epsilon and 1 - epsilon < p[2] < 1 + epsilon)]
if len(vertex_group_ransac.planes) == 0:
return
cell_complex = CellComplex(vertex_group_ransac.planes, vertex_group_ransac.bounds,
vertex_group_ransac.points_grouped,
build_graph=True, additional_planes=additional_planes,
initial_bound=[[-0.5, -0.5, -0.5], [0.5, 0.5, 0.5]], quiet=True)
cell_complex.refine_planes() # additional planes are not refined
cell_complex.prioritise_planes()
cell_complex.construct()
cell_complex.save_obj(filepath=Path(cc_output_path).with_suffix('.obj'))
cell_complex.save(filepath=cc_output_path)
def multi_merge_vg2cc(vg_ransac_dir, vg_city3d_dir, cc_output_dir):
"""
Merge vertex groups from RANSAC and from City3D, then generate complexes, with multiprocessing.
"""
args = []
num_workers = 42
vg_filenames_ransac = glob.glob(vg_ransac_dir + '/*.vg')
for vg_filename_ransac in vg_filenames_ransac:
stem = Path(vg_filename_ransac).stem
vg_filenames_city3d = (Path(vg_city3d_dir) / stem).with_suffix('.vg')
if not vg_filenames_city3d.exists():
continue
cc_filenames_output = (Path(cc_output_dir) / stem).with_suffix('.cc')
if cc_filenames_output.exists():
continue
args.append([vg_filename_ransac, vg_filenames_city3d, cc_filenames_output])
with multiprocessing.Pool(processes=num_workers) as pool:
# call with multiprocessing
for _ in tqdm(pool.imap_unordered(merge_vg2cc, args), desc='Creating complexes from vertex groups',
total=len(args)):
pass
def vg2cc(args):
"""
Create cell complex from vertex group.
"""
vg_path, cc_path = args
# print(vg_path)
vertex_group = VertexGroup(filepath=vg_path, refit=False, global_group=True)
cell_complex = CellComplex(vertex_group.planes, vertex_group.bounds, vertex_group.points_grouped,
build_graph=True, additional_planes=None,
initial_bound=[[-0.5, -0.5, -0.5], [0.5, 0.5, 0.5]])
cell_complex.refine_planes(theta=5 * 3.1416 / 180, epsilon=0.002)
cell_complex.construct()
cell_complex.save_obj(filepath=Path(cc_path).with_suffix('.obj'))
cell_complex.save(filepath=cc_path)
def multi_vg2cc(vg_dir, cc_dir):
"""
Create cell complexes from vertex groups with multiprocessing.
"""
args = []
num_workers = 38
vg_filenames = glob.glob(vg_dir + '/*.vg')
for vg_filename in vg_filenames:
stem = Path(vg_filename).stem
cc_filename = (Path(cc_dir) / stem).with_suffix('.cc')
args.append([vg_filename, cc_filename])
with multiprocessing.Pool(processes=num_workers) as pool:
# call with multiprocessing
for _ in tqdm(pool.imap_unordered(vg2cc, args), desc='Creating complexes from vertex groups', total=len(args)):
pass
def coordinate2index(x, reso, coord_type='2d'):
""" Generate grid index of points
Args:
x (tensor): points (normalized to [0, 1])
reso (int): defined resolution
coord_type (str): coordinate type
"""
x = (x * reso).long()
if coord_type == '2d': # plane
index = x[:, :, 0] + reso * x[:, :, 1] # [B, N, 1]
index = index[:, None, :] # [B, 1, N]
return index
def normalize_coordinate(p, padding=0, plane='xz', scale=1.0):
""" Normalize coordinate to [0, 1] for unit cube experiments
Args:
p (tensor): point
padding (float): conventional padding parameter of ONet for unit cube, so [-0.5, 0.5] -> [-0.55, 0.55]
plane (str): plane feature type, ['xz', 'xy', 'yz']
scale: normalize scale
"""
if plane == 'xz':
xy = p[:, :, [0, 2]]
elif plane == 'xy':
xy = p[:, :, [0, 1]]
else:
xy = p[:, :, [1, 2]]
xy_new = xy / (1 + padding + 10e-6) # (-0.5, 0.5)
xy_new = xy_new + 0.5 # range (0, 1)
# f there are outliers out of the range
if xy_new.max() >= 1:
xy_new[xy_new >= 1] = 1 - 10e-6
if xy_new.min() < 0:
xy_new[xy_new < 0] = 0.0
return xy_new
def normalize_3d_coordinate(p, padding=0):
""" Normalize coordinate to [0, 1] for unit cube experiments.
Corresponds to our 3D model
Args:
p (tensor): point
padding (float): conventional padding paramter of ONet for unit cube, so [-0.5, 0.5] -> [-0.55, 0.55]
"""
p_nor = p / (1 + padding + 10e-4) # (-0.5, 0.5)
p_nor = p_nor + 0.5 # range (0, 1)
# f there are outliers out of the range
if p_nor.max() >= 1:
p_nor[p_nor >= 1] = 1 - 10e-4
if p_nor.min() < 0:
p_nor[p_nor < 0] = 0.0
return p_nor
class map2local(object):
"""
Add new keys to the given input
Args:
s (float): the defined voxel size
pos_encoding (str): method for the positional encoding, linear|sin_cos
"""
def __init__(self, s, pos_encoding='linear'):
super().__init__()
self.s = s
self.pe = positional_encoding(basis_function=pos_encoding)
def __call__(self, p):
p = torch.remainder(p, self.s) / self.s # always positive
# p = torch.fmod(p, self.s) / self.s # same sign as input p!
p = self.pe(p)
return p
class positional_encoding(object):
""" Positional Encoding (presented in NeRF)
Args:
basis_function (str): basis function
"""
def __init__(self, basis_function='sin_cos'):
super().__init__()
self.func = basis_function
L = 10
freq_bands = 2. ** (np.linspace(0, L - 1, L))
self.freq_bands = freq_bands * math.pi
def __call__(self, p):
if self.func == 'sin_cos':
out = []
p = 2.0 * p - 1.0 # chagne to the range [-1, 1]
for freq in self.freq_bands:
out.append(torch.sin(freq * p))
out.append(torch.cos(freq * p))
p = torch.cat(out, dim=2)
return p
def make_3d_grid(bb_min, bb_max, shape):
"""
Makes a 3D grid.
Args:
bb_min (tuple): bounding box minimum
bb_max (tuple): bounding box maximum
shape (tuple): output shape
"""
size = shape[0] * shape[1] * shape[2]
pxs = torch.linspace(bb_min[0], bb_max[0], int(shape[0]))
pys = torch.linspace(bb_min[1], bb_max[1], int(shape[1]))
pzs = torch.linspace(bb_min[2], bb_max[2], int(shape[2]))
pxs = pxs.view(-1, 1, 1).expand(*shape).contiguous().view(size)
pys = pys.view(1, -1, 1).expand(*shape).contiguous().view(size)
pzs = pzs.view(1, 1, -1).expand(*shape).contiguous().view(size)
p = torch.stack([pxs, pys, pzs], dim=1)
return p
if __name__ == '__main__':
pass