-
Notifications
You must be signed in to change notification settings - Fork 5
/
dataset.py
237 lines (226 loc) · 8.39 KB
/
dataset.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
import open3d as o3d
import torch
import h5py
import numpy as np
import torch.utils.data as data
import torchvision.transforms as transforms
import os
import random
#from utils import *
def read_points(filename, dataset):
if dataset == 'suncg' or dataset == 'fusion':
pcd = o3d.read_point_cloud(filename)
coord = torch.from_numpy(np.array(pcd.points)).float()
color = torch.from_numpy(np.array(pcd.colors)).float()
return coord, color
elif dataset == 'shapenet':
hash_tab = {
'all': {
'name': 'Test',
'label': 100,
'emd1': 0.0,
'emd2': 0.0,
'emd3': 0.0,
'cd1': 0.0,
'cd2': 0.0,
'cd3': 0.0,
'cnt': 0
},
'04530566': {
'name': 'Watercraft',
'label': 1,
'emd1': 0.0,
'emd2': 0.0,
'emd3': 0.0,
'cd1': 0.0,
'cd2': 0.0,
'cd3': 0.0,
'cnt': 0
},
'02933112': {
'name': 'Cabinet',
'label': 2,
'emd1': 0.0,
'emd2': 0.0,
'emd3': 0.0,
'cd1': 0.0,
'cd2': 0.0,
'cd3': 0.0,
'cnt': 0
},
'04379243': {
'name': 'Table',
'label': 3,
'emd1': 0.0,
'emd2': 0.0,
'emd3': 0.0,
'cd1': 0.0,
'cd2': 0.0,
'cd3': 0.0,
'cnt': 0
},
'02691156': {
'name': 'Airplane',
'label': 4,
'emd1': 0.0,
'emd2': 0.0,
'emd3': 0.0,
'cd1': 0.0,
'cd2': 0.0,
'cd3': 0.0,
'cnt': 0
},
'02958343': {
'name': 'Car',
'label': 5,
'emd1': 0.0,
'emd2': 0.0,
'emd3': 0.0,
'cd1': 0.0,
'cd2': 0.0,
'cd3': 0.0,
'cnt': 0
},
'03001627': {
'name': 'Chair',
'label': 6,
'emd1': 0.0,
'emd2': 0.0,
'emd3': 0.0,
'cd1': 0.0,
'cd2': 0.0,
'cd3': 0.0,
'cnt': 0
},
'04256520': {
'name': 'Couch',
'label': 7,
'emd1': 0.0,
'emd2': 0.0,
'emd3': 0.0,
'cd1': 0.0,
'cd2': 0.0,
'cd3': 0.0,
'cnt': 0
},
'03636649': {
'name': 'Lamp',
'label': 8,
'emd1': 0.0,
'emd2': 0.0,
'emd3': 0.0,
'cd1': 0.0,
'cd2': 0.0,
'cd3': 0.0,
'cnt': 0
}
}
fh5 = h5py.File(filename, 'r')
label = float(hash_tab[filename.split("/")[-2]]['label'])
coord = torch.from_numpy(np.array(fh5['data'])).float()
color = torch.from_numpy(
np.ones_like(np.array(fh5['data'])) / 11 * label).float()
return coord, color
def resample_pcd(pcd, n):
"""Drop or duplicate points so that pcd has exactly n points"""
idx = np.random.permutation(pcd.shape[0])
if idx.shape[0] < n:
idx = np.concatenate(
[idx, np.random.randint(pcd.shape[0], size=n - pcd.shape[0])])
return pcd[idx[:n]], idx[:n]
class ShapeNet(data.Dataset):
def __init__(self, train=True, npoints=2048, dataset_name='shapenet'):
self.dataset = dataset_name
if train:
if self.dataset == 'suncg':
self.list_path = './data/train_suncg.list'
elif self.dataset == 'fusion':
self.list_path = './data/train_fusion.list'
elif self.dataset == 'shapenet':
self.list_path = './data/train_shapenet.list'
else:
if self.dataset == 'suncg':
self.list_path = './data/valid_suncg.list'
elif self.dataset == 'fusion':
self.list_path = './data/test_fusion.list'
elif self.dataset == 'shapenet':
self.list_path = './data/valid_shapenet.list'
self.npoints = npoints
self.train = train
with open(os.path.join(self.list_path)) as file:
self.model_list = [line.strip().replace('/', '/') for line in file]
random.shuffle(self.model_list)
self.len = len(self.model_list)
def __getitem__(self, index):
model_id = self.model_list[index]
scan_id = index
if self.train:
if self.dataset == 'suncg':
part, part_color = read_points(
os.path.join(
"/media/wangyida/HDD/database/SUNCG_Yida/train/pcd_partial/",
'%s.pcd' % model_id), self.dataset)
comp, comp_color = read_points(
os.path.join(
"/media/wangyida/HDD/database/SUNCG_Yida/train/pcd_complete/",
'%s.pcd' % model_id), self.dataset)
if self.dataset == 'fusion':
part, part_color = read_points(
os.path.join(
"/media/wangyida/HDD/database/050_200/train/pcd_partial/",
'%s.pcd' % model_id), self.dataset)
comp, comp_color = read_points(
os.path.join(
"/media/wangyida/HDD/database/050_200/train/pcd_complete/",
'%s.pcd' % model_id), self.dataset)
elif self.dataset == 'shapenet':
part, part_color = read_points(
os.path.join(
"/media/wangyida/HDD/database/shapenet/train/partial/",
'%s.h5' % model_id), self.dataset)
comp, comp_color = read_points(
os.path.join(
"/media/wangyida/HDD/database/shapenet/train/gt/",
'%s.h5' % model_id), self.dataset)
else:
if self.dataset == 'suncg':
part, part_color = read_points(
os.path.join(
"/media/wangyida/HDD/database/SUNCG_Yida/test/pcd_partial/",
'%s.pcd' % model_id))
comp, comp_color = read_points(
os.path.join(
"/media/wangyida/HDD/database/SUNCG_Yida/test/pcd_complete/",
'%s.pcd' % model_id), self.dataset)
elif self.dataset == 'fusion':
part, part_color = read_points(
os.path.join(
"/media/wangyida/HDD/database/050_200/test/pcd_partial/",
'%s.pcd' % model_id))
comp, comp_color = read_points(
os.path.join(
"/media/wangyida/HDD/database/050_200/test/pcd_complete/",
'%s.pcd' % model_id), self.dataset)
elif self.dataset == 'shapenet':
part, part_color = read_points(
os.path.join(
"/media/wangyida/HDD/database/shapenet/val/partial/",
'%s.h5' % model_id), self.dataset)
comp, comp_color = read_points(
os.path.join(
"/media/wangyida/HDD/database/shapenet/val/gt/",
'%s.h5' % model_id), self.dataset)
part_sampled, idx_sampled = resample_pcd(part, self.npoints)
part_seg = np.round(part_color[idx_sampled] * 11)
comp_sampled, idx_sampled = resample_pcd(comp, self.npoints)
# comp_sampled, idx_sampled = resample_pcd(comp, self.npoints * 8)
comp_seg = np.round(comp_color[idx_sampled] * 11)
"""
comp_seg = []
for i in range (1, 12):
import ipdb; ipdb.set_trace()
comp_seg.append(resample_pcd(comp_sampled[comp_color == i], 512))
"""
return model_id, part_sampled, comp_sampled, part_seg, comp_seg
def __len__(self):
return self.len