-
Notifications
You must be signed in to change notification settings - Fork 0
/
datasets.py
428 lines (370 loc) · 16.1 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
import cv2
import torch.utils.data as data
import torch
import os
import os.path
import numpy as np
import sys
import json
class IndoorPointRecord(object):
def __init__(self, path):
self.path = path
def call_input(self, mode, yaw):
return os.path.join(self.path, 'mode=%s_%d.png'%(mode, yaw))
@property
def OverviewMask(self):
return os.path.join(self.path, 'OverviewMask.png')
@property
def OverviewMap(self):
return os.path.join(self.path, 'topdown_view.png')
@property
def video_frame(self):
return os.path.join(self.path, 'topdown_video_frame.png')
class RealRGBRecord(object):
def __init__(self, path):
self.path = path
def call_input(self, mode, yaw):
# print(self.path)
return os.path.join(self.path, '%s-%d.png'%(mode, yaw))
@property
def OverviewMask(self):
return os.path.join(self.path, 'topdown-semantics.png')
@property
def OverviewRGB(self):
return os.path.join(self.path, 'topdown-rgb_filled.png')
class CarPointRecord(object):
def __init__(self, path, height):
self.path = path
self.episode, self.camera, self.number = self.path.split('/')
self.height = height
def RGB_FRONT_input(self):
return os.path.join(self.episode, 'CAM_RGB_FRONT', self.number)
def RGB_FRONT_RIGHT_input(self):
return os.path.join(self.episode, 'CAM_RGB_FRONT_RIGHT', self.number)
def RGB_FRONT_LEFT_input(self):
return os.path.join(self.episode, 'CAM_RGB_FRONT_LEFT', self.number)
def RGB_BACK_RIGHT_input(self):
return os.path.join(self.episode, 'CAM_RGB_BACK_RIGHT', self.number)
def RGB_BACK_LEFT_input(self):
return os.path.join(self.episode, 'CAM_RGB_BACK_LEFT', self.number)
def RGB_BACK_input(self):
return os.path.join(self.episode, 'CAM_RGB_BACK', self.number)
@property
def OverviewMask(self):
return os.path.join(self.episode, 'CAM_SemSeg_TOPDOWN_{}'.format(self.height), self.number)
@property
def OverviewMap(self):
return os.path.join(self.episode, 'CAM_RGB_TOPDOWN_{}'.format(self.height), self.number)
@property
def video_frame(self):
return os.path.join(self.path, 'topdown_video_frame.png')
@property
def get_number(self):
return self.number.split('.')[0]
class OVMDataset(data.Dataset):
def __init__(self, datadir, split, transform, is_train=True,
num_views=8, input_size=224, label_size=25,
use_mask=False, use_depth=False):
self.datadir = datadir
self.split = split
self.transform = transform
self.is_train = is_train
self.num_views = num_views
self.input_size = input_size
self.label_size = label_size
self.use_mask = use_mask
self.use_depth= use_depth
print('use mask: ', self.use_mask)
print('use depth: ', self.use_depth)
self._parse_list()
self._parse_color()
def _parse_list(self):
tmp = []
for x in open(self.split):
x = x.strip()
if x.split('/')[0] != 'e042c74158a0b1dad5f1b6a689fd056a':
tmp.append(x)
self.coor_list = [IndoorPointRecord(item) for item in tmp]
print('Coordinate number:%d'%(len(self.coor_list)))
def _parse_color(self):
with open('./metadata/colormap_coarse.csv') as f:
lines = f.readlines()
cat = []
for line in lines:
line = line.rstrip()
cat.append(line)
cat = cat[1:]
self.label_dic = {}
for i, value in enumerate(cat):
key = ','.join(value.split(',')[1:])
self.label_dic[key] = i
def __getitem__(self, item):
example = self.coor_list[item]
input_data = list()
gap = int(8/self.num_views)
if self.use_mask:
input_img = cv2.imread(os.path.join(self.datadir, example.call_input('sem', 0)))
elif self.use_depth:
input_img = cv2.imread(os.path.join(self.datadir, example.call_input('depth', 0)))
else:
input_img = cv2.imread(os.path.join(self.datadir, example.call_input('rgb', 0)))
for i, rank in enumerate([2, 2, 3, 3, 0, 0, 1, 1]):
if i % gap == 0:
split_data = input_img[:, input_img.shape[1]//4*rank:input_img.shape[1]//4*(rank+1)]
split_data = cv2.resize(split_data, (self.input_size, self.input_size), interpolation=cv2.INTER_NEAREST)
input_data.extend([split_data])
if self.num_views > 4:
if self.use_mask:
input_img_45 = cv2.imread(os.path.join(self.datadir, example.call_input('sem', 45)))
elif self.use_depth:
input_img_45 = cv2.imread(os.path.join(self.datadir, example.call_input('depth', 45)))
else:
input_img_45 = cv2.imread(os.path.join(self.datadir, example.call_input('rgb', 45)))
for i, rank in enumerate([2, 3, 0, 1]):
split_data = input_img_45[:, input_img_45.shape[1]//4*rank:input_img_45.shape[1]//4*(rank+1)]
split_data = cv2.resize(split_data, (self.input_size, self.input_size), interpolation=cv2.INTER_NEAREST)
input_data[2*i+1] = split_data
image = input_data.copy()
input_data = self.transform(input_data)
om_orig = cv2.imread(os.path.join(self.datadir, example.OverviewMask))
om = cv2.resize(om_orig[:, :, ::-1], (self.label_size, self.label_size), interpolation=cv2.INTER_NEAREST)
mask = np.uint8(np.zeros((self.label_size, self.label_size)))
for i, _ in enumerate(om):
for j, _ in enumerate(om[0]):
key = ','.join([str(x) for x in om[i, j]])
mask[i, j] = self.label_dic[key]
mask = torch.from_numpy(mask)
if not self.is_train:
rgb_origin = np.concatenate([np.expand_dims(x, 0) for x in image], axis=0)
return input_data, mask.long(), rgb_origin, om_orig
return input_data, mask.long()
def __len__(self):
return len(self.coor_list)
class House3D_Dataset(data.Dataset):
def __init__(self, datadir, split, transform, is_train=True,
num_views=8, input_size=224, label_size=128):
self.datadir = datadir
self.split = split
self.transform = transform
self.is_train = is_train
self.num_views = num_views
self.input_size = input_size
self.label_size = label_size
self._parse_list()
self._parse_color()
def _parse_list(self):
tmp = []
for x in open(self.split):
x = x.strip()
if x.split('/')[0] != 'e042c74158a0b1dad5f1b6a689fd056a':
tmp.append(x)
self.coor_list = [IndoorPointRecord(item) for item in tmp]
self.coor_list = self.coor_list[:15800]
print('Coordinate number:%d'%(len(self.coor_list)))
def _parse_color(self):
with open('./metadata/colormap_coarse.csv') as f:
lines = f.readlines()
cat = []
for line in lines:
line = line.rstrip()
cat.append(line)
cat = cat[1:]
self.label_dic = {}
for i, value in enumerate(cat):
key = ','.join(value.split(',')[1:])
self.label_dic[key] = i
def __getitem__(self, item):
example = self.coor_list[item]
input_data = list()
gap = int(8/self.num_views)
input_img = cv2.imread(os.path.join(self.datadir, example.call_input('rgb', 0)))
for i, rank in enumerate([2, 2, 3, 3, 0, 0, 1, 1]):
if i % gap == 0:
split_data = input_img[:, input_img.shape[1]//4*rank:input_img.shape[1]//4*(rank+1)]
split_data = cv2.resize(split_data, (self.input_size, self.input_size), interpolation=cv2.INTER_NEAREST)
input_data.extend([split_data])
if self.num_views > 4:
input_img_45 = cv2.imread(os.path.join(self.datadir, example.call_input('rgb', 45)))
for i, rank in enumerate([2, 3, 0, 1]):
split_data = input_img_45[:, input_img_45.shape[1]//4*rank:input_img_45.shape[1]//4*(rank+1)]
split_data = cv2.resize(split_data, (self.input_size, self.input_size), interpolation=cv2.INTER_NEAREST)
input_data[2*i+1] = split_data
image = input_data.copy()
input_data = self.transform(input_data)
om_orig = cv2.imread(os.path.join(self.datadir, example.OverviewMask))
om = cv2.resize(om_orig[:, :, ::-1], (self.label_size, self.label_size), interpolation=cv2.INTER_NEAREST)
mask = np.uint8(np.zeros((self.label_size, self.label_size)))
for i, _ in enumerate(om):
for j, _ in enumerate(om[0]):
key = ','.join([str(x) for x in om[i, j]])
mask[i, j] = self.label_dic[key]
mask = torch.from_numpy(mask)
if not self.is_train:
rgb_origin = np.concatenate([np.expand_dims(x, 0) for x in image], axis=0)
return input_data, mask.long(), rgb_origin, om_orig
return input_data, mask.long()
def __len__(self):
return len(self.coor_list)
class MP3D_Dataset(data.Dataset):
def __init__(self, datadir, split, transform, is_train=True,
num_views=8, input_size=224, label_size=25):
self.datadir = datadir
self.split = split
self.transform = transform
self.is_train = is_train
self.num_views = num_views
self.input_size = input_size
self.label_size = label_size
self._parse_list()
self._parse_color()
def _parse_list(self):
tmp = []
for x in open(self.split):
x = x.strip()
tmp.append(x)
self.coor_list = [RealRGBRecord(item) for item in tmp]
self.coor_list = self.coor_list[:220]
print('Coordinate number:%d'%(len(self.coor_list)))
def _parse_color(self):
with open('./metadata/colormap_coarse.csv') as f:
lines = f.readlines()
cat = []
for line in lines:
line = line.rstrip()
cat.append(line)
cat = cat[1:]
self.label_dic = {}
for i, value in enumerate(cat):
key = ','.join(value.split(',')[1:])
self.label_dic[key] = i
def __getitem__(self, item):
example = self.coor_list[item]
input_data = list()
gap = int(8 / self.num_views)
for yaw in [gap * i * 45 for i in range(self.num_views)]:
rgb_img = cv2.imread(os.path.join(self.datadir, example.call_input('rgb_filled', yaw)))
rgb_img = cv2.resize(rgb_img, (self.input_size, self.input_size), interpolation=cv2.INTER_NEAREST)
input_data.extend([rgb_img[:, :, ::-1]])
input_data = self.transform(input_data)
return input_data
def __len__(self):
return len(self.coor_list)
class Carla_Dataset(data.Dataset):
def __init__(self, datadir, split, transform, is_train=True,
num_views=8, input_size=224, label_size=25,
use_mask=False, use_depth=False, height=10):
self.datadir = datadir
self.split = split
self.transform = transform
self.is_train = is_train
self.num_views = num_views
self.input_size = input_size
self.label_size = label_size
self.use_mask = use_mask
self.use_depth = use_depth
self.height = height
print('use mask: ', self.use_mask)
print('use depth: ', self.use_depth)
self._parse_list()
self._parse_color()
def _parse_list(self):
tmp = []
for x in open(self.split):
x = x.strip()
tmp.append(x)
self.coor_list = [CarPointRecord(item, height=self.height) for item in tmp]
print('Coordinate number:%d' % (len(self.coor_list)))
def _parse_color(self):
with open('./metadata/colormap_coarse.csv') as f:
lines = f.readlines()
cat = []
for line in lines:
line = line.rstrip()
cat.append(line)
cat = cat[1:]
self.label_dic = {}
for i, value in enumerate(cat):
key = ','.join(value.split(',')[1:])
self.label_dic[key] = i
def __getitem__(self, item):
example = self.coor_list[item]
input_data = list()
input_img = [cv2.imread(os.path.join(self.datadir, example.RGB_FRONT_input())),
cv2.imread(os.path.join(self.datadir, example.RGB_FRONT_LEFT_input())),
cv2.imread(os.path.join(self.datadir, example.RGB_FRONT_RIGHT_input())),
cv2.imread(os.path.join(self.datadir, example.RGB_BACK_input())),
cv2.imread(os.path.join(self.datadir, example.RGB_BACK_RIGHT_input())),
cv2.imread(os.path.join(self.datadir, example.RGB_BACK_LEFT_input()))]
split_data = []
for img in input_img:
resized_img = cv2.resize(img, (self.input_size, self.input_size), interpolation=cv2.INTER_NEAREST)
split_data.append(resized_img)
input_data.extend(split_data)
image = input_data.copy()
input_data = self.transform(input_data)
om_orig = cv2.imread(os.path.join(self.datadir, example.OverviewMask))
om = cv2.resize(om_orig[:, :, ::-1], (self.label_size, self.label_size), interpolation=cv2.INTER_NEAREST)
mask = torch.from_numpy(om[:, :, 0])
if not self.is_train:
rgb_origin = np.concatenate([np.expand_dims(x, 0) for x in image], axis=0)
return input_data, mask.long(), rgb_origin, om_orig[:, :, -1]
return input_data, mask.long()
def __len__(self):
return len(self.coor_list)
class nuScenes_Dataset(data.Dataset):
def __init__(self, datadir, split, transform, is_train=True,
num_views=8, input_size=224, label_size=25,
use_mask=False, use_depth=False, height=10):
self.datadir = datadir
self.split = split
self.transform = transform
self.is_train = is_train
self.num_views = num_views
self.input_size = input_size
self.label_size = label_size
self.use_mask = use_mask
self.use_depth = use_depth
self.height = height
print('use mask: ', self.use_mask)
print('use depth: ', self.use_depth)
self._parse_list()
self._parse_color()
def _parse_list(self):
tmp = []
for x in open(self.split):
x = x.strip()
tmp.append(x)
self.coor_list = [CarPointRecord(item, height=self.height) for item in tmp]
print('Coordinate number:%d' % (len(self.coor_list)))
def _parse_color(self):
with open('./metadata/colormap_coarse.csv') as f:
lines = f.readlines()
cat = []
for line in lines:
line = line.rstrip()
cat.append(line)
cat = cat[1:]
self.label_dic = {}
for i, value in enumerate(cat):
key = ','.join(value.split(',')[1:])
self.label_dic[key] = i
def __getitem__(self, item):
example = self.coor_list[item]
input_data = list()
input_img = [cv2.imread(os.path.join(self.datadir, example.RGB_FRONT_input())),
cv2.imread(os.path.join(self.datadir, example.RGB_FRONT_LEFT_input())),
cv2.imread(os.path.join(self.datadir, example.RGB_FRONT_RIGHT_input())),
cv2.imread(os.path.join(self.datadir, example.RGB_BACK_input())),
cv2.imread(os.path.join(self.datadir, example.RGB_BACK_RIGHT_input())),
cv2.imread(os.path.join(self.datadir, example.RGB_BACK_LEFT_input()))]
sys.stdout.flush()
split_data = []
for img in input_img:
# print(img is None)
resized_img = cv2.resize(img, (self.input_size, self.input_size), interpolation=cv2.INTER_NEAREST)
split_data.append(resized_img)
input_data.extend(split_data)
input_data = self.transform(input_data)
return input_data
def __len__(self):
return len(self.coor_list)