-
Notifications
You must be signed in to change notification settings - Fork 3
/
dataset_3d.py
667 lines (562 loc) · 27.3 KB
/
dataset_3d.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
'''
* Copyright (c) 2023, salesforce.com, inc.
* All rights reserved.
* SPDX-License-Identifier: BSD-3-Clause
* For full license text, see the LICENSE file in the repo root or https://opensource.org/licenses/BSD-3-Clause
'''
import random
import torch
import numpy as np
import torch.utils.data as data
import csv
import yaml
from easydict import EasyDict
from utils.io import IO
from utils.build import DATASETS
from utils.logger import *
from utils.build import build_dataset_from_cfg
import json
from tqdm import tqdm
import pickle
from PIL import Image
def pil_loader(path):
# open path as file to avoid ResourceWarning (https://github.com/python-pillow/Pillow/issues/835)
with open(path, 'rb') as f:
img = Image.open(f)
return img.convert('RGB')
def pc_normalize(pc):
centroid = np.mean(pc, axis=0)
pc = pc - centroid
m = np.max(np.sqrt(np.sum(pc**2, axis=1)))
pc = pc / m
return pc
def farthest_point_sample(point, npoint):
"""
Input:
xyz: pointcloud data, [N, D]
npoint: number of samples
Return:
centroids: sampled pointcloud index, [npoint, D]
"""
N, D = point.shape
xyz = point[:,:3]
centroids = np.zeros((npoint,))
distance = np.ones((N,)) * 1e10
farthest = np.random.randint(0, N)
for i in range(npoint):
centroids[i] = farthest
centroid = xyz[farthest, :]
dist = np.sum((xyz - centroid) ** 2, -1)
mask = dist < distance
distance[mask] = dist[mask]
farthest = np.argmax(distance, -1)
point = point[centroids.astype(np.int32)]
return point
def rotate_point_cloud(batch_data):
""" Randomly rotate the point clouds to augument the dataset
rotation is per shape based along up direction
Input:
BxNx3 array, original batch of point clouds
Return:
BxNx3 array, rotated batch of point clouds
"""
rotated_data = np.zeros(batch_data.shape, dtype=np.float32)
for k in range(batch_data.shape[0]):
rotation_angle = np.random.uniform() * 2 * np.pi
cosval = np.cos(rotation_angle)
sinval = np.sin(rotation_angle)
rotation_matrix = np.array([[cosval, 0, sinval],
[0, 1, 0],
[-sinval, 0, cosval]])
shape_pc = batch_data[k, ...]
rotated_data[k, ...] = np.dot(shape_pc.reshape((-1, 3)), rotation_matrix)
return rotated_data
def random_point_dropout(batch_pc, max_dropout_ratio=0.875):
''' batch_pc: BxNx3 '''
for b in range(batch_pc.shape[0]):
dropout_ratio = np.random.random()*max_dropout_ratio # 0~0.875
drop_idx = np.where(np.random.random((batch_pc.shape[1]))<=dropout_ratio)[0]
if len(drop_idx)>0:
batch_pc[b,drop_idx,:] = batch_pc[b,0,:] # set to the first point
return batch_pc
def random_scale_point_cloud(batch_data, scale_low=0.8, scale_high=1.25):
""" Randomly scale the point cloud. Scale is per point cloud.
Input:
BxNx3 array, original batch of point clouds
Return:
BxNx3 array, scaled batch of point clouds
"""
B, N, C = batch_data.shape
scales = np.random.uniform(scale_low, scale_high, B)
for batch_index in range(B):
batch_data[batch_index,:,:] *= scales[batch_index]
return batch_data
def shift_point_cloud(batch_data, shift_range=0.1):
""" Randomly shift point cloud. Shift is per point cloud.
Input:
BxNx3 array, original batch of point clouds
Return:
BxNx3 array, shifted batch of point clouds
"""
B, N, C = batch_data.shape
shifts = np.random.uniform(-shift_range, shift_range, (B,3))
for batch_index in range(B):
batch_data[batch_index,:,:] += shifts[batch_index,:]
return batch_data
def jitter_point_cloud(batch_data, sigma=0.01, clip=0.05):
""" Randomly jitter points. jittering is per point.
Input:
BxNx3 array, original batch of point clouds
Return:
BxNx3 array, jittered batch of point clouds
"""
B, N, C = batch_data.shape
assert(clip > 0)
jittered_data = np.clip(sigma * np.random.randn(B, N, C), -1*clip, clip)
jittered_data += batch_data
return jittered_data
def rotate_perturbation_point_cloud(batch_data, angle_sigma=0.06, angle_clip=0.18):
""" Randomly perturb the point clouds by small rotations
Input:
BxNx3 array, original batch of point clouds
Return:
BxNx3 array, rotated batch of point clouds
"""
rotated_data = np.zeros(batch_data.shape, dtype=np.float32)
for k in range(batch_data.shape[0]):
angles = np.clip(angle_sigma*np.random.randn(3), -angle_clip, angle_clip)
Rx = np.array([[1,0,0],
[0,np.cos(angles[0]),-np.sin(angles[0])],
[0,np.sin(angles[0]),np.cos(angles[0])]])
Ry = np.array([[np.cos(angles[1]),0,np.sin(angles[1])],
[0,1,0],
[-np.sin(angles[1]),0,np.cos(angles[1])]])
Rz = np.array([[np.cos(angles[2]),-np.sin(angles[2]),0],
[np.sin(angles[2]),np.cos(angles[2]),0],
[0,0,1]])
R = np.dot(Rz, np.dot(Ry,Rx))
shape_pc = batch_data[k, ...]
rotated_data[k, ...] = np.dot(shape_pc.reshape((-1, 3)), R)
return rotated_data
import os, sys, h5py
BASE_DIR = os.path.dirname(os.path.abspath(__file__))
sys.path.append(BASE_DIR)
@DATASETS.register_module()
class ScanObjectNN(data.Dataset):
def __init__(self, config):
super().__init__()
self.root = config.DATA_PATH
h5 = h5py.File(os.path.join(self.root, 'main_split_nobg', 'test_objectdataset.h5'), 'r')
self.points = np.array(h5['data']).astype(np.float32)
self.labels = np.array(h5['label']).astype(int)
h5.close()
with open(os.path.join(self.root, 'ScanObjectNN_shape_names.txt'), 'r') as f:
lines = f.readlines()
lines = [line.rstrip() for line in lines]
self.shape_name=lines
def __getitem__(self, idx):
pt_idxs = np.arange(0, self.points.shape[1]) # 2048
current_points = self.points[idx, pt_idxs].copy()
current_points = torch.from_numpy(current_points).float()
label = self.labels[idx]
return current_points, label, self.shape_name[int(label)]
def __len__(self):
return self.points.shape[0]
@DATASETS.register_module()
class ModelNet(data.Dataset):
def __init__(self, config):
self.root = config.DATA_PATH
self.npoints = config.npoints
self.use_normals = config.USE_NORMALS
self.num_category = config.NUM_CATEGORY
self.process_data = True
self.uniform = True
self.generate_from_raw_data = False
split = config.subset
self.subset = config.subset
self.sets = 'Hard'
if self.num_category == 10:
self.catfile = os.path.join(self.root, 'modelnet10_shape_names.txt')
else:
self.catfile = os.path.join(self.root, 'modelnet40_shape_names.txt')
self.cat = [line.rstrip() for line in open(self.catfile)]
self.classes = dict(zip(self.cat, range(len(self.cat))))
shape_ids = {}
if self.num_category == 10:
shape_ids['train'] = [line.rstrip() for line in open(os.path.join(self.root, 'modelnet10_train.txt'))]
shape_ids['test'] = [line.rstrip() for line in open(os.path.join(self.root, 'modelnet10_test.txt'))]
else:
shape_ids['train'] = [line.rstrip() for line in open(os.path.join(self.root, 'modelnet40_train.txt'))]
shape_ids['test'] = [line.rstrip() for line in open(os.path.join(self.root, 'modelnet40_test.txt'))]
assert (split == 'train' or split == 'test')
shape_names = ['_'.join(x.split('_')[0:-1]) for x in shape_ids[split]]
self.datapath = [(shape_names[i], os.path.join(self.root, shape_names[i], shape_ids[split][i]) + '.txt') for i
in range(len(shape_ids[split]))]
print_log('The size of %s data is %d' % (split, len(self.datapath)), logger='ModelNet')
if self.uniform:
self.save_path = os.path.join(self.root,
'modelnet%d_%s_%dpts_fps.dat' % (self.num_category, split, self.npoints))
else:
self.save_path = os.path.join(self.root,
'modelnet%d_%s_%dpts.dat' % (self.num_category, split, self.npoints))
if self.process_data:
if not os.path.exists(self.save_path):
# make sure you have raw data in the path before you enable generate_from_raw_data=True.
if self.generate_from_raw_data:
print_log('Processing data %s (only running in the first time)...' % self.save_path, logger='ModelNet')
self.list_of_points = [None] * len(self.datapath)
self.list_of_labels = [None] * len(self.datapath)
for index in tqdm(range(len(self.datapath)), total=len(self.datapath)):
fn = self.datapath[index]
cls = self.classes[self.datapath[index][0]]
cls = np.array([cls]).astype(np.int32)
point_set = np.loadtxt(fn[1], delimiter=',').astype(np.float32)
if self.uniform:
point_set = farthest_point_sample(point_set, self.npoints)
print_log("uniformly sampled out {} points".format(self.npoints))
else:
point_set = point_set[0:self.npoints, :]
self.list_of_points[index] = point_set
self.list_of_labels[index] = cls
with open(self.save_path, 'wb') as f:
pickle.dump([self.list_of_points, self.list_of_labels], f)
else:
# no pre-processed dataset found and no raw data found, then load 8192 points dataset then do fps after.
self.save_path = os.path.join(self.root,
'modelnet%d_%s_%dpts_fps.dat' % (
self.num_category, split, 8192))
print_log('Load processed data from %s...' % self.save_path, logger='ModelNet')
print_log('since no exact points pre-processed dataset found and no raw data found, load 8192 pointd dataset first, then do fps to {} after, the speed is excepted to be slower due to fps...'.format(self.npoints), logger='ModelNet')
with open(self.save_path, 'rb') as f:
self.list_of_points, self.list_of_labels = pickle.load(f)
else:
print_log('Load processed data from %s...' % self.save_path, logger='ModelNet')
with open(self.save_path, 'rb') as f:
self.list_of_points, self.list_of_labels = pickle.load(f)
self.shape_names_addr = os.path.join(self.root, 'modelnet40_shape_names.txt')
with open(self.shape_names_addr) as file:
lines = file.readlines()
lines = [line.rstrip() for line in lines]
self.shape_names = lines
# TODO: disable for backbones except for PointNEXT!!!
self.use_height = config.use_height
# The split of hard_set and medium_set, you can change it as your wish.
self.hard_set = ['cone', 'curtain', 'door', 'dresser', 'glass_box', 'mantel', 'night_stand', 'person', 'plant', 'radio', 'range_hood', 'sink', 'stairs', 'tent', 'toilet', 'tv_stand', 'xbox']
self.medium_set = ['cone', 'cup', 'curtain', 'door', 'dresser', 'glass_box', 'mantel', 'monitor', 'night_stand', 'person', 'plant', 'radio', 'range_hood', 'sink', 'stairs', 'stool', 'tent', 'toilet', 'tv_stand', 'vase', 'wardrobe', 'xbox']
def __len__(self):
return len(self.list_of_labels)
def _get_item(self, index):
if self.process_data:
point_set, label = self.list_of_points[index], self.list_of_labels[index]
else:
fn = self.datapath[index]
cls = self.classes[self.datapath[index][0]]
label = np.array([cls]).astype(np.int32)
point_set = np.loadtxt(fn[1], delimiter=',').astype(np.float32)
if self.uniform:
point_set = farthest_point_sample(point_set, self.npoints)
else:
point_set = point_set[0:self.npoints, :]
if self.npoints < point_set.shape[0]:
point_set = farthest_point_sample(point_set, self.npoints)
point_set[:, 0:3] = pc_normalize(point_set[:, 0:3])
if not self.use_normals:
point_set = point_set[:, 0:3]
if self.use_height:
self.gravity_dim = 1
height_array = point_set[:, self.gravity_dim:self.gravity_dim + 1] - point_set[:,
self.gravity_dim:self.gravity_dim + 1].min()
point_set = np.concatenate((point_set, height_array), axis=1)
return point_set, label[0]
def __getitem__(self, index):
points, label = self._get_item(index)
pt_idxs = np.arange(0, points.shape[0]) # 2048
if self.subset == 'train':
np.random.shuffle(pt_idxs)
current_points = points[pt_idxs].copy()
current_points = torch.from_numpy(current_points).float()
label_name = self.shape_names[int(label)]
if self.sets == 'Medium':
if not label_name in self.medium_set:
return None, None, None
label = self.medium_set.index(label_name)
elif self.sets == 'Hard':
if not label_name in self.hard_set:
return None, None, None
label = self.hard_set.index(label_name)
return current_points, label, label_name
@DATASETS.register_module()
class ShapeNet(data.Dataset):
def __init__(self, config):
self.data_root = config.DATA_PATH
self.pc_path = config.PC_PATH
self.subset = config.subset
self.npoints = config.npoints
self.tokenizer = config.tokenizer
self.train_transform = config.train_transform
self.id_map_addr = os.path.join(config.DATA_PATH, 'taxonomy.json')
self.rendered_image_addr = config.IMAGE_PATH
self.picked_image_type = ['', '_depth0001']
self.picked_rotation_degrees = list(range(0, 360, 12))
self.picked_rotation_degrees = [(3 - len(str(degree))) * '0' + str(degree) if len(str(degree)) < 3 else str(degree) for degree in self.picked_rotation_degrees]
self.sub_id_map_addr = "./data/ShapeNet_all.csv"
with open(self.sub_id_map_addr, 'r') as f:
sub_id_sets = list(csv.reader(f))
self.sub_id_map = {}
for _, _, subSynsetId, modelId, _ in sub_id_sets:
self.sub_id_map[modelId] = subSynsetId
with open(self.id_map_addr, 'r') as f:
self.id_map = json.load(f)
with open('./data/ShapeNet_Class2Labels.json', 'r') as f:
self.label_map = json.load(f)
self.prompt_template_addr = os.path.join('./data/templates.json')
with open(self.prompt_template_addr) as f:
self.templates = json.load(f)[config.pretrain_dataset_prompt]
self.synset_id_map = {}
for id_dict in self.id_map:
synset_id = id_dict["synsetId"]
self.synset_id_map[synset_id] = id_dict
self.data_list_file = os.path.join(self.data_root, f'{self.subset}.txt')
test_data_list_file = os.path.join(self.data_root, 'test.txt')
self.sample_points_num = self.npoints
self.whole = config.get('whole')
print_log(f'[DATASET] sample out {self.sample_points_num} points', logger='ShapeNet-55')
print_log(f'[DATASET] Open file {self.data_list_file}', logger='ShapeNet-55')
with open(self.data_list_file, 'r') as f:
lines = f.readlines()
if self.whole:
with open(test_data_list_file, 'r') as f:
test_lines = f.readlines()
print_log(f'[DATASET] Open file {test_data_list_file}', logger='ShapeNet-55')
lines = test_lines + lines
self.file_list = []
for line in lines:
line = line.strip()
taxonomy_id = line.split('-')[0]
model_id = line[len(taxonomy_id) + 1:].split('.')[0]
sub_taxonomy_id = self.sub_id_map[model_id].zfill(8) if model_id in self.sub_id_map else taxonomy_id
self.file_list.append({
'taxonomy_id': taxonomy_id,
'sub_taxonomy_id': sub_taxonomy_id,
'model_id': model_id,
'file_path': line
})
print_log(f'[DATASET] {len(self.file_list)} instances were loaded', logger='ShapeNet-55')
self.permutation = np.arange(self.npoints)
self.args = config.args
self.uniform = True
self.augment = True
self.all_image = config.args.all_image
# =================================================
# TODO: disable for backbones except for PointNEXT!!!
self.use_height = config.use_height
# =================================================
if self.augment:
print("using augmented point clouds.")
# self.template = "a point cloud model of {}."
def pc_norm(self, pc):
""" pc: NxC, return NxC """
centroid = np.mean(pc, axis=0)
pc = pc - centroid
m = np.max(np.sqrt(np.sum(pc ** 2, axis=1)))
pc = pc / m
return pc
def random_sample(self, pc, num):
np.random.shuffle(self.permutation)
pc = pc[self.permutation[:num]]
return pc
def __getitem__(self, idx):
sample = self.file_list[idx]
data = IO.get(os.path.join(self.pc_path, sample['file_path'])).astype(np.float32)
if self.uniform and self.sample_points_num < data.shape[0]:
data = farthest_point_sample(data, self.sample_points_num)
else:
data = self.random_sample(data, self.sample_points_num)
data = self.pc_norm(data)
if self.augment:
data = random_point_dropout(data[None, ...])
data = random_scale_point_cloud(data)
data = shift_point_cloud(data)
data = rotate_perturbation_point_cloud(data)
data = rotate_point_cloud(data)
data = data.squeeze()
if self.use_height:
self.gravity_dim = 1
height_array = data[:, self.gravity_dim:self.gravity_dim + 1] - data[:,
self.gravity_dim:self.gravity_dim + 1].min()
data = np.concatenate((data, height_array), axis=1)
data = torch.from_numpy(data).float()
else:
data = torch.from_numpy(data).float()
if self.args.constrative_text == 'super':
captions = self.synset_id_map[sample['taxonomy_id']]['name']
else:
captions = self.synset_id_map[sample['sub_taxonomy_id']]['name']
captions = [caption.strip() for caption in captions.split(',') if caption.strip()]
caption = random.choice(captions)
captions = []
tokenized_captions = []
for template in self.templates:
captions.append(template.format(caption))
# tokenized_captions.append(self.tokenizer(template.format(caption)))
tokenized_captions.append(self.tokenizer(caption))
tokenized_captions = torch.stack(tokenized_captions)
picked_model_rendered_image_addr = self.rendered_image_addr + '/' +\
sample['taxonomy_id'] + '-' + sample['model_id'] + '/'
if self.all_image:
images = []
all_degress = []
for image_type in self.picked_image_type:
degrees = random.sample(range(30), 1)
# offsets = random.sample([-3, -2, -1, 1, 2, 3], 1)
# for offset in offsets:
# degrees.append((degrees[0] + offset) % 30)
all_degress.extend(degrees)
for degree in degrees:
picked_image_name = sample['taxonomy_id'] + '-' + sample['model_id'] + '_r_' +\
str(self.picked_rotation_degrees[degree]) +\
image_type + '.png'
picked_image_addr = picked_model_rendered_image_addr + picked_image_name
try:
image = pil_loader(picked_image_addr)
image = self.train_transform(image)
images.append(image)
except:
raise ValueError("image is corrupted: {}".format(picked_image_addr))
images = torch.stack(images, dim=0)
all_degress = torch.from_numpy(np.array(all_degress))
else:
picked_image_name = sample['taxonomy_id'] + '-' + sample['model_id'] + '_r_' +\
str(random.choice(self.picked_rotation_degrees)) +\
random.choice(self.picked_image_type) + '.png'
picked_image_addr = picked_model_rendered_image_addr + picked_image_name
degrees = None
try:
image = pil_loader(picked_image_addr)
image = self.train_transform(image)
except:
raise ValueError("image is corrupted: {}".format(picked_image_addr))
if self.args.class_text == 'None':
label_map = self.label_map['super']
else:
label_map = self.label_map[self.args.class_text]
return_image = image
return_label = label_map[sample['taxonomy_id']]
if self.all_image:
return_image = images
else:
all_degress = 0
if self.args.class_text == 'sub':
return_label = label_map[sample['sub_taxonomy_id']]
return return_label, sample['model_id'], tokenized_captions, data, return_image, all_degress
def __len__(self):
return len(self.file_list)
import collections.abc as container_abcs
int_classes = int
from torch._six import string_classes
import re
default_collate_err_msg_format = (
"default_collate: batch must contain tensors, numpy arrays, numbers, "
"dicts or lists; found {}")
np_str_obj_array_pattern = re.compile(r'[SaUO]')
def customized_collate_fn(batch):
r"""Puts each data field into a tensor with outer dimension batch size"""
elem = batch[0]
elem_type = type(elem)
if isinstance(batch, list):
batch = [example for example in batch if example[4] is not None]
if isinstance(elem, torch.Tensor):
out = None
if torch.utils.data.get_worker_info() is not None:
# If we're in a background process, concatenate directly into a
# shared memory tensor to avoid an extra copy
numel = sum([x.numel() for x in batch])
storage = elem.storage()._new_shared(numel)
out = elem.new(storage)
return torch.stack(batch, 0, out=out)
elif elem_type.__module__ == 'numpy' and elem_type.__name__ != 'str_' \
and elem_type.__name__ != 'string_':
if elem_type.__name__ == 'ndarray' or elem_type.__name__ == 'memmap':
# array of string classes and object
if np_str_obj_array_pattern.search(elem.dtype.str) is not None:
raise TypeError(default_collate_err_msg_format.format(elem.dtype))
return customized_collate_fn([torch.as_tensor(b) for b in batch])
elif elem.shape == (): # scalars
return torch.as_tensor(batch)
elif isinstance(elem, float):
return torch.tensor(batch, dtype=torch.float64)
elif isinstance(elem, int_classes):
return torch.tensor(batch)
elif isinstance(elem, string_classes):
return batch
elif isinstance(elem, container_abcs.Mapping):
return {key: customized_collate_fn([d[key] for d in batch]) for key in elem}
elif isinstance(elem, tuple) and hasattr(elem, '_fields'): # namedtuple
return elem_type(*(customized_collate_fn(samples) for samples in zip(*batch)))
elif isinstance(elem, container_abcs.Sequence):
# check to make sure that the elements in batch have consistent size
it = iter(batch)
elem_size = len(next(it))
if not all(len(elem) == elem_size for elem in it):
raise RuntimeError('each element in list of batch should be of equal size')
transposed = zip(*batch)
return [customized_collate_fn(samples) for samples in transposed]
raise TypeError(default_collate_err_msg_format.format(elem_type))
def zs_collate_fn(batch):
batch = list(filter(lambda x:x[0] is not None, batch))
if len(batch) == 0:
return None, None, None
return torch.utils.data.dataloader.default_collate(batch)
def merge_new_config(config, new_config):
for key, val in new_config.items():
if not isinstance(val, dict):
if key == '_base_':
with open(new_config['_base_'], 'r') as f:
try:
val = yaml.load(f, Loader=yaml.FullLoader)
except:
val = yaml.load(f)
config[key] = EasyDict()
merge_new_config(config[key], val)
else:
config[key] = val
continue
if key not in config:
config[key] = EasyDict()
merge_new_config(config[key], val)
return config
def cfg_from_yaml_file(cfg_file):
config = EasyDict()
with open(cfg_file, 'r') as f:
new_config = yaml.load(f, Loader=yaml.FullLoader)
merge_new_config(config=config, new_config=new_config)
return config
class Dataset_3D():
def __init__(self, args, tokenizer, dataset_type, train_transform=None):
if dataset_type == 'train':
self.dataset_name = args.pretrain_dataset_name
elif dataset_type == 'val':
self.dataset_name = args.validate_dataset_name
else:
raise ValueError("not supported dataset type.")
with open('./data/dataset_catalog.json', 'r') as f:
self.dataset_catalog = json.load(f)
self.dataset_usage = self.dataset_catalog[self.dataset_name]['usage']
self.dataset_split = self.dataset_catalog[self.dataset_name][self.dataset_usage]
self.dataset_config_dir = self.dataset_catalog[self.dataset_name]['config']
self.tokenizer = tokenizer
self.train_transform = train_transform
self.pretrain_dataset_prompt = args.pretrain_dataset_prompt
self.validate_dataset_prompt = args.validate_dataset_prompt
self.build_3d_dataset(args, self.dataset_config_dir)
def build_3d_dataset(self, args, config):
config = cfg_from_yaml_file(config)
config.tokenizer = self.tokenizer
config.train_transform = self.train_transform
config.pretrain_dataset_prompt = self.pretrain_dataset_prompt
config.validate_dataset_prompt = self.validate_dataset_prompt
config.args = args
config.use_height = args.use_height
config.npoints = args.npoints
config_others = EasyDict({'subset': self.dataset_split, 'whole': True})
self.dataset = build_dataset_from_cfg(config, config_others)