-
Notifications
You must be signed in to change notification settings - Fork 0
/
dataset.py
executable file
·175 lines (157 loc) · 6.7 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
import os
import numpy as np
import torch
from torch.utils.data import Dataset
import matplotlib
import random
matplotlib.use('Agg')
from pyntcloud import PyntCloud
import h5py
from pytorch3d.ops.knn import knn_gather, knn_points
class ShapeNet_Completion_Seg(Dataset):
def __init__(self, dataset_path, mode='train', point_num=256, category='chair', transform=None, shift=False, rotation=False,
scaling=False, drop=False, mask_input=None , gt_num=16384):
super().__init__()
self.transform = transform
self.shift = shift
self.rotation = rotation
self.scaling = scaling
self.drop = drop
self.point_num = point_num
categories_all = {
"airplane": "02691156",
"car": "02958343",
"chair": "03001627",
"lamp": "03636649",
"table": "04379243",
}
self.PART_NUM = {
"airplane": 4,
"bag": 2,
"cap": 2,
"car": 4,
"chair": 4,
"earphone": 3,
"guitar": 3,
"knife": 2,
"lamp": 4,
"laptop": 2,
"motorbike": 6,
"mug": 2,
"pistol": 3,
"rocket": 3,
"skateboard": 3,
"table": 3,
}
PART_MIN = {
"airplane": 0,
"bag": 4,
"cap": 6,
"car": 8,
"chair": 12,
"earphone": 16,
"guitar": 19,
"knife": 22,
"lamp": 24,
"laptop": 28,
"motorbike": 30,
"mug": 36,
"pistol": 38,
"rocket": 41,
"skateboard": 44,
"table": 47,
}
if category:
if category not in categories_all.keys():
raise Exception('Categoty not found !')
self.categories = [category]
else:
self.categories = categories_all.keys()
self.categories = list(self.categories)
print(self.categories)
self.path_label_pairs = []
for category in self.categories:
label = list(categories_all.keys()).index(category)
min_part = PART_MIN[category]
folder_path = os.path.join(dataset_path ,mode)
object_path = [os.path.join(folder_path, categories_all[category] ,'{}'.format(name)) for name in sorted(os.listdir(os.path.join(folder_path,categories_all[category] ))) if name != '.DS_Store']
for path in object_path[:]:
partial_list = [f for f in os.listdir(os.path.join(path, 'partial')) if f.endswith('.npy')]
for part_id in partial_list:
pair = (path, os.path.join(path,'partial',part_id), label, min_part)
self.path_label_pairs.append(pair)
print('mode:', mode)
print('number of data', len(self.path_label_pairs))
self.part_nums = self.PART_NUM[category]
self.mask_input = mask_input
def __len__(self):
return len(self.path_label_pairs)
def __getitem__(self, index):
#############################
# token 0 = padding
# token 1 = mask
# other tokens = original seg_label + 2
#############################
obj_path, partial_path ,label, min_part = self.path_label_pairs[index]
label = torch.LongTensor([label])
gt_point_seg_full = np.load(obj_path + '/gt_with_seg.npy')
gt_seg_full = torch.FloatTensor(gt_point_seg_full[:, -1]) - min_part
gt_index = torch.randperm(16384)
gt_point_seg = torch.FloatTensor(gt_point_seg_full[gt_index][:16384, :])
gt_point, gt_seg = gt_point_seg[:,:3], gt_point_seg[:,-1] - min_part
gt_token = gt_seg + 2
# count part ratio for gt input
gt_part_count = torch.unique(gt_seg_full, return_counts=True)
gt_part_ratio = torch.zeros(self.part_nums)
for i in range(len(gt_part_count[0])):
gt_part_ratio[gt_part_count[0].long()[i]] = gt_part_count[1][i]
gt_part_ratio = gt_part_ratio / 16384
partial_point = np.load(partial_path)
partial_point = torch.FloatTensor(partial_point)
n, _ = partial_point.size()
if n == 0:
print(partial_path)
return None
if n < self.point_num:
padding = torch.zeros((self.point_num-n, 4))
padding[:, -1] -= (2 - min_part)
partial_point = torch.cat((partial_point, padding))
elif n > self.point_num:
sample_index = torch.randperm(n)
partial_point = partial_point[sample_index]
partial_point = partial_point[:self.point_num,:]
n = self.point_num
partial_point, partial_seg = partial_point[:, :3], partial_point[:, -1] - min_part
partial_token = partial_seg + 2
# count part ratio for partial input
partial_part_count = torch.unique(partial_seg, return_counts=True)
partial_part_ratio = torch.zeros(self.part_nums)
if -2 in partial_part_count[0]:
partial_part_count = (partial_part_count[0][1:], partial_part_count[1][1:])
for i in range(len(partial_part_count[0])):
partial_part_ratio[partial_part_count[0].long()[i]] = partial_part_count[1][i]
partial_part_ratio = partial_part_ratio.float() / partial_part_count[1].sum()
if self.mask_input:
for i in range(len(partial_token)):
prob = random.random()
if partial_token[i] != 0 and (prob>self.mask_input):
partial_token[i] = 1
special_token = torch.zeros((self.part_nums,3))
special_token_label = torch.arange(self.part_nums).float()
#############################
# Return data and label #
# partial_point #
# partial_seg #
# partial_token #
# gt_point #
# gt_seg #
# gt_token #
# label #
# part_nums #
# gt_part_ratio #
# partial_part_ratio #
#############################
return torch.cat((special_token,partial_point)), torch.cat((special_token_label,partial_seg)), torch.cat((special_token_label+2,partial_token)), \
torch.cat((special_token,gt_point)), torch.cat((special_token_label,gt_seg)), torch.cat((special_token_label+2,gt_token)), label, n+self.part_nums, gt_part_ratio, partial_part_ratio
if __name__ == "__main__":
shapenet_set = ShapeNet_Completion_Seg(dataset_path='../../shapenet_completion_full_with_part', mode='test',category='lamp')