-
Notifications
You must be signed in to change notification settings - Fork 19
/
dataset.py
212 lines (193 loc) · 10.1 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
import os
import sys
BASE_DIR = os.path.dirname(os.path.abspath(__file__))
sys.path.append(os.path.join(BASE_DIR, 'utils'))
sys.path.append(os.path.join(BASE_DIR, 'tf_ops/sampling'))
import copy
import io_util
import tensorflow as tf
import tf_sampling
import numpy as np
import scipy.io as sio
class ScanNetDataset():
def __init__(self, src_mesh_path, src_label_path, list_path, cache_path, npoint=18000, npoint_ins=512, is_augment=False, permute_points=True):
'''
src_mesh_path: path to ScanNet mesh
src_label_path: path to ScanNet label
list_path: a plain txt file containing all the file names
cache_path: path to the cached files
npoint: number of sampled points per scene
npoint_ins: number of sampled points per instance
'''
self.npoint = npoint
self.npoint_ins = npoint_ins
self.ngroup = 0
self.is_augment = is_augment
self.permute_points = permute_points
self.file_list = io_util.read_txt(list_path)
self.data_list = {}
if os.path.exists(cache_path):
#### collect the processed scene data
self.data_list = np.load(cache_path)['data_list'].item()
#### collect the number of instances per scene
self.ngroup = np.load(cache_path)['ngroup'].item()
else:
self.cache_file(src_mesh_path, src_label_path, cache_path)
def cache_file(self, src_mesh_path, src_label_path, cache_path):
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
with tf.device('/gpu:0'):
pc_tf = tf.placeholder(tf.float32)
ind_tf = tf_sampling.farthest_point_sample(self.npoint, pc_tf)
ind_ins_tf = tf_sampling.farthest_point_sample(self.npoint_ins, pc_tf)
sess = tf.Session(config=config)
nfile = len(self.file_list)
#### valid class ids defined in ScanNet
VALID_CLASS_IDS = np.array([3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 16, 24, 28, 33, 34, 36, 39])
target_sem_idx = np.arange(40)
count = 0
for i in range(40):
if i in VALID_CLASS_IDS:
count += 1
target_sem_idx[i] = count
else:
target_sem_idx[i] = 0
for index in range(nfile):
print(np.float32(index)/nfile)
curpc_color = io_util.read_color_ply(os.path.join(src_mesh_path, self.file_list[index]+'.ply'))
curpc = curpc_color[:,:3].astype(np.float32)
curcolor = curpc_color[:,3:].astype(np.float32)/255.0
curgroup = io_util.read_label_txt(os.path.join(src_label_path, 'group_'+self.file_list[index]+'.txt')).astype(np.int32)
curseg = io_util.read_label_txt(os.path.join(src_label_path, 'sem_'+self.file_list[index]+'.txt')).astype(np.int32)
curseg[curseg>=40] = 0
curseg[curseg<0] = 0
curseg = self.changem(curseg, np.arange(40), target_sem_idx)
curngroup = np.max(curgroup)+1
#### collect instance information
valid_group_indicator = np.zeros(curngroup)
target_idx = np.zeros(1+curngroup)
count = 0
for i in range(curngroup):
if np.sum(curgroup==i)==0:
target_idx[i+1] = 0
elif np.round(np.mean(curseg[curgroup==i])).astype('int32')!=0:
valid_group_indicator[i] = 1
count += 1
target_idx[i+1] = count
else:
target_idx[i+1] = 0
curgroup = self.changem(curgroup, np.arange(-1, curngroup), target_idx).astype('int32')
curgroup[curgroup<0] = 0
curngroup = 1+np.sum(valid_group_indicator).astype('int32') # group zero is background
#### resample each scene to a fix number of points
if curngroup>self.ngroup:
self.ngroup = curngroup
if self.npoint<curpc.shape[0]:
choice = sess.run(ind_tf, feed_dict={pc_tf: np.expand_dims(curpc,0)})[0]
pc = curpc[choice,:]
color = curcolor[choice,:]
group_label = curgroup[choice]
seg_label = curseg[choice]
elif self.npoint==curpc.shape[1]:
pc = copy.deepcopy(curpc)
color = copy.deepcopy(curcolor)
group_label = copy.deepcopy(curgroup)
seg_label = copy.deepcopy(curseg)
else:
choice = np.random.choice(curpc.shape[0], self.npoint - curpc.shape[0])
pc = np.concatenate((curpc,curpc[choice,:]), 0)
color = np.concatenate((curcolor,curcolor[choice,:]), 0)
group_label = np.concatenate((curgroup, curgroup[choice]), 0)
seg_label = np.concatenate((curseg, curseg[choice]), 0)
#### resample each instance to a fix number of points
pc_ins = np.zeros((curngroup, self.npoint_ins, 3), dtype=np.float32)
for j in range(1,curngroup):
curins = curpc[curgroup==j,:]
if self.npoint_ins<curins.shape[0]:
choice = sess.run(ind_ins_tf, feed_dict={pc_tf: np.expand_dims(curins,0)})[0]
pc_ins[j,:,:] = curins[choice,:]
elif self.npoint_ins==curins.shape[0]:
pc_ins[j,:,:] = copy.deepcopy(curins)
else:
choice = np.random.choice(curins.shape[0], self.npoint_ins - curins.shape[0])
pc_ins[j,:,:] = np.concatenate((curins,curins[choice,:]), 0)
#### data tuple for each scene
#### group_label indicates instance label
#### seg_label indicates semantic label
self.data_list[index] = (pc, color, group_label, seg_label, pc_ins, curngroup)
np.savez_compressed(cache_path, data_list=self.data_list, ngroup=self.ngroup)
def gen_rotation_matrix(self):
rotation_angle = np.random.uniform() * 2 * np.pi
cosval = np.cos(rotation_angle)
sinval = np.sin(rotation_angle)
rotation_matrix = np.array([[cosval, sinval, 0],
[-sinval, cosval, 0],
[0, 0, 1]])
return rotation_matrix
def changem(self, input_array, source_idx, target_idx):
mapping = {}
for i, sidx in enumerate(source_idx):
mapping[sidx] = target_idx[i]
input_array = np.array([mapping[i] for i in input_array])
return input_array
def __getitem__(self, index):
'''
Return:
pc: [npoint, 3], point cloud of the whole scene, in world coord system
color: [npoint, 3], rgb for each point
pc_ins_full: [ngroup, npoint_ins, 3], point cloud of all instances, in world coord system, all zero for background instance
group_label: [npoint], instance label, 0 means background
group_indicator: [ngroup], indicates which groups are valid, usually the first few
seg_label: [npoint], semantic label, 0 means background class
'''
pc, color, group_label, seg_label, pc_ins, curngroup = self.data_list[index]
#### randomly permute the point order
if self.permute_points:
ridx = np.random.permutation(pc.shape[0])
pc = copy.deepcopy(pc[ridx,:])
color = copy.deepcopy(color[ridx,:])
seg_label = copy.deepcopy(seg_label[ridx])
group_label = copy.deepcopy(group_label[ridx])
else:
pc = copy.deepcopy(pc)
color = copy.deepcopy(color)
seg_label = copy.deepcopy(seg_label)
group_label = copy.deepcopy(group_label)
#### pad instance to a maximum instance number
group_indicator = np.zeros((self.ngroup), dtype=np.int32)
group_indicator[:curngroup] = 1
pc_ins_full = np.zeros((self.ngroup, self.npoint_ins, 3), dtype=np.float32)
pc_ins_full[:curngroup,:,:] = pc_ins
#### augmenting the input scene with random rotation around z-axis and random translation
if self.is_augment:
R = self.gen_rotation_matrix()
pc = np.matmul(pc, R)
pc_ins_full = np.reshape(np.matmul(np.reshape(pc_ins_full, [-1, 3]), R), pc_ins_full.shape)
# aug translation
t = np.random.normal(0,1,[1,3])
pc += t
pc_ins_full += np.reshape(t, [1,1,3])
#### compute axis aligned bounding box for all instances
bbox_ins_full = np.zeros((self.ngroup, 6), dtype=np.float32)
bbox_ins_full[:, :3] = (np.max(pc_ins_full,1)+np.min(pc_ins_full,1))/2
bbox_ins_full[:, 3:] = np.max(pc_ins_full,1)-np.min(pc_ins_full,1)
return pc, color, pc_ins_full, group_label, group_indicator, seg_label, bbox_ins_full
def __len__(self):
return len(self.file_list)
if __name__ == '__main__':
npoint = 18000 # number of sampled points per scene
npoint_ins = 512 # number of sampled points per instance
if not os.path.exists(os.path.join(BASE_DIR, 'data/cache')):
os.makedirs(os.path.join(BASE_DIR, 'data/cache'))
src_mesh_path = os.path.join(BASE_DIR, 'data/scannet_preprocessed/mesh/scans')
src_label_path = os.path.join(BASE_DIR, 'data/scannet_preprocessed/label/scans')
train_list = os.path.join(BASE_DIR, 'data/scannet/scannet_train.txt')
val_list = os.path.join(BASE_DIR, 'data/scannet/scannet_val.txt')
train_cache = os.path.join(BASE_DIR, 'data/cache/train_%d_%d.npz'%(npoint, npoint_ins))
val_cache = os.path.join(BASE_DIR, 'data/cache/val_%d_%d.npz'%(npoint, npoint_ins))
trainDataset = ScanNetDataset(src_mesh_path, src_label_path, train_list, train_cache, npoint=npoint, npoint_ins=npoint_ins, is_augment=True)
print(len(trainDataset))
pc, color, pc_ins_full, group_label, group_indicator, seg_label, bbox_ins_full = trainDataset[0]
valDataset = ScanNetDataset(src_mesh_path, src_label_path, val_list, val_cache, npoint=npoint, npoint_ins=npoint_ins, is_augment=False)
print(len(valDataset))
pc_val, color_val, pc_ins_full_val, group_label_val, group_indicator_val, seg_label_val, bbox_ins_full_val = valDataset[0]