/
dataset_stats.py
268 lines (220 loc) · 10.7 KB
/
dataset_stats.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
#
# Performing Dataset Statistics Analysis
# Yuhang Ming
#
import pickle
import numpy as np
from os import makedirs
from os.path import join, exists, isfile
from utils.ply import read_ply
if __name__ == '__main__':
"""
Number of pts:
task <5000 5000-6000 6000-7000 7000-8000 8000-9000 >9000
training [6089 2383 2409 2467 2459 19295]
[0.1734659 0.06788787 0.06862857 0.0702809 0.07005299 0.54968378]
validation [1710 655 615 608 588 5517]
[0.17641597 0.06757454 0.06344785 0.06272568 0.06066233 0.56917363]
test [629 224 207 208 214 2126]
[0.17433481 0.06208426 0.05737251 0.05764967 0.05931264 0.58924612]
"""
# set dir
base_dir = '/media/yohann/fastStorage/data/'
scannet_path = '/media/yohann/fastStorage/data/ScanNetPR'
data_split_path = join(scannet_path, 'tools/Tasks/Benchmark')
scannet_data_path = join(scannet_path, 'scans')
tasks = ['training', 'validation', 'test']
# load all valid pcds for all tasks
valid_pcd_file = join(scannet_path, 'VLAD_triplets', 'vlad_pcd.txt')
with open(valid_pcd_file, "rb") as f:
# dict: key = scene string, val = list of filenames
all_scene_pcds = pickle.load(f)
########################
## Statistic analysis ##
########################
stats = []
num = {'training': 35102, 'validation': 9693, 'test': 3608}
count = []
dict_pcd_size_ind = {}
dict_pcd_size = {}
# Loop through all tasks
for task in tasks:
print(task)
# get scene names to process
if task == 'training':
scene_file_name = join(data_split_path, 'scannetv2_train.txt')
scenes = np.sort(np.loadtxt(scene_file_name, dtype=str))
elif task == 'validation':
scene_file_name = join(data_split_path, 'scannetv2_val.txt')
scenes = np.sort(np.loadtxt(scene_file_name, dtype=str))
else:
scene_file_name = join(data_split_path, 'scannetv2_test.txt')
scenes = np.loadtxt(scene_file_name, dtype=str)
# print(' ', str(len(scenes)), 'sequences')
task_pcd_size_ind = {'<5000': [], '5000-10000': [], '10000-15000': [], '>15000': []}
task_stats = np.array([0,0,0,0])
task_count = 0
# loop through all scenes
for i, scene in enumerate(scenes):
num_scene_pcd = len(all_scene_pcds[scene])
scene_pcd_size = []
# Loop through pcds
for j, pcd in enumerate(all_scene_pcds[scene]):
print('{}%'.format(int(100*task_count/num[task])), flush=True, end='\r')
# # Get frame id
# actual_frame_id = int(pcd[13:-8])
# Get pcd file path
pcd_file_path = join(scannet_data_path, scene, 'input_pcd_0mean', pcd)
# load file
data = read_ply(pcd_file_path)
pts = np.vstack((data['x'], data['y'], data['z'])).astype(np.float32).T
num_pts = pts.shape[0]
scene_pcd_size.append(num_pts)
if num_pts < 5000:
task_stats[0] += 1
task_pcd_size_ind['<5000'].append((i,j))
elif num_pts < 10000:
task_stats[1] += 1
task_pcd_size_ind['5000-10000'].append((i,j))
elif num_pts < 15000:
task_stats[2] += 1
task_pcd_size_ind['10000-15000'].append((i,j))
else:
task_stats[3] += 1
task_pcd_size_ind['>15000'].append((i,j))
task_count += 1
dict_pcd_size[scene] = scene_pcd_size
dict_pcd_size_ind[task] = task_pcd_size_ind
print('100%')
stats.append(task_stats)
count.append(task_count)
print(task_stats)
print(task_stats/task_count)
print(len(dict_pcd_size_ind['training']['<5000']))
print(len(dict_pcd_size_ind['validation']['<5000']))
print(len(dict_pcd_size_ind['test']['<5000']))
# print(dict_pcd_size['scene0000_00'])
# pcd_size_file = join(scannet_path, 'VLAD_triplets', 'pcd_size.txt')
# with open(pcd_size_file, "wb") as f:
# pickle.dump(dict_pcd_size, f)
# pcd_size_ind_file = join(scannet_path, 'VLAD_triplets', 'pcd_size_ind.txt')
# with open(pcd_size_ind_file, "wb") as f:
# pickle.dump(dict_pcd_size_ind, f)
print(stats)
# ##########################
# ## Dataset manipulation ##
# ##########################
# import shutil
# scannet_pr_path = join(base_dir, 'ScanNetPR')
# if not exists(scannet_pr_path):
# makedirs(scannet_pr_path)
# pcd_path = join(scannet_pr_path, 'input_pcd')
# # Retrive All relevant pcds and RGB-D images and store it in a new directory
# # Loop though all PR pcds
# count = 0
# max_count = len(all_scene_pcds)
# for scene, scene_pcds in all_scene_pcds.items():
# # print(scene, '\n')
# # print('{}%'.format(int(100*count/max_count)), flush=True, end='\r')
# # new scene folder
# new_scene_path = join(scannet_pr_path, scene)
# # print(new_scene_path)
# if not exists(new_scene_path):
# makedirs(new_scene_path)
# # loop through all pcds in the scene
# # store pcds, rgb, depth, pose, intrinsics in corresponding folder
# pre_input_pcd_path = join(scannet_data_path, 'input_pcd', scene)
# new_input_pcd_path = join(new_scene_path, 'input_pcd')
# if not exists(new_input_pcd_path):
# makedirs(new_input_pcd_path)
# pre_input_pcd_0mean_path = join(scannet_data_path, 'input_pcd_0mean', scene)
# new_input_pcd_0mean_path = join(new_scene_path, 'input_pcd_0mean')
# if not exists(new_input_pcd_0mean_path):
# makedirs(new_input_pcd_0mean_path)
# pre_pnvlad_pcd_path = join(scannet_data_path, 'pnvlad_pcd', scene)
# new_pnvlad_pcd_path = join(new_scene_path, 'pnvlad_pcd')
# if not exists(new_pnvlad_pcd_path):
# makedirs(new_pnvlad_pcd_path)
# pre_color_path = join(scannet_data_path, scene, 'color')
# new_color_path = join(new_scene_path, 'color')
# if not exists(new_color_path):
# makedirs(new_color_path)
# pre_depth_path = join(scannet_data_path, scene, 'depth')
# new_depth_path = join(new_scene_path, 'depth')
# if not exists(new_depth_path):
# makedirs(new_depth_path)
# pre_pose_path = join(scannet_data_path, scene, 'pose')
# new_pose_path = join(new_scene_path, 'pose')
# if not exists(new_pose_path):
# makedirs(new_pose_path)
# pre_info_file = join(scannet_data_path, scene, scene+'.txt')
# new_info_file = join(new_scene_path, scene+'.txt')
# if isfile(pre_info_file):
# shutil.copy(pre_info_file, new_info_file)
# max_pcd = len(scene_pcds)
# for j, pcd in enumerate(scene_pcds):
# actual_frame_id = str(int(pcd[13:-8]))
# print('{:.2f}% - {:.2f}%'.format(float(100*count/max_count), float(100*j/max_pcd)), flush=True, end='\r')
# # print(' ', actual_frame_id, pcd)
# # get file dirs
# src_files = []
# dst_files = []
# # pcds
# src_files.append(join(pre_input_pcd_path, pcd))
# dst_files.append(join(new_input_pcd_path, pcd))
# src_files.append(join(pre_input_pcd_0mean_path, pcd))
# dst_files.append(join(new_input_pcd_0mean_path, pcd))
# src_files.append(join(pre_pnvlad_pcd_path, pcd[:-3]+'bin'))
# dst_files.append(join(new_pnvlad_pcd_path, pcd[:-3]+'bin'))
# # color, depth, pose
# src_files.append(join(pre_color_path, actual_frame_id+'.jpg'))
# dst_files.append(join(new_color_path, actual_frame_id+'.jpg'))
# src_files.append(join(pre_depth_path, actual_frame_id+'.png'))
# dst_files.append(join(new_depth_path, actual_frame_id+'.png'))
# src_files.append(join(pre_pose_path, actual_frame_id+'.txt'))
# dst_files.append(join(new_pose_path, actual_frame_id+'.txt'))
# # copy
# for i in range(len(src_files)):
# if isfile(src_files[i]):
# shutil.copy(src_files[i], dst_files[i])
# count += 1
# # print('{:.2f}% - 100\%'.format(float(100*count/max_count)), flush=True, end='\r')
# # # test pcd files
# # data = read_ply('/media/adam/Datasets/datasets/ScanNetPR/scene0000_00/input_pcd/scene0000_00_0_sub.ply')
# # pts = np.vstack((data['x'], data['y'], data['z'])).astype(np.float32).T
# # num_pts = pts.shape[0]
# # mean = np.mean(pts, axis=0)
# # print(mean)
# # data = read_ply('/media/adam/Datasets/datasets/ScanNetPR/scene0000_00/input_pcd_0mean/scene0000_00_0_sub.ply')
# # pts = np.vstack((data['x'], data['y'], data['z'])).astype(np.float32).T
# # num_pts = pts.shape[0]
# # mean = np.mean(pts, axis=0)
# # print(mean)
# # Loop through all tasks
# for task in tasks:
# print(task)
# # get scene names to process
# if task == 'training':
# scene_file_name = join(data_split_path, 'scannetv2_train.txt')
# scenes = np.sort(np.loadtxt(scene_file_name, dtype=str))
# elif task == 'validation':
# scene_file_name = join(data_split_path, 'scannetv2_val.txt')
# scenes = np.sort(np.loadtxt(scene_file_name, dtype=str))
# else:
# scene_file_name = join(data_split_path, 'scannetv2_test.txt')
# scenes = np.loadtxt(scene_file_name, dtype=str)
# # print(' ', str(len(scenes)), 'sequences')
# task_pcd_size_ind = {'<5000': [], '5000-7000': [], '7000-9000': [], '>9000': []}
# task_stats = np.array([0,0,0,0])
# task_count = 0
# # loop through all scenes
# for i, scene in enumerate(scenes):
# num_scene_pcd = len(all_scene_pcds[scene])
# scene_pcd_size = []
# # Loop through pcds
# for j, pcd in enumerate(all_scene_pcds[scene]):
# print('{}%'.format(int(100*task_count/num[task])), flush=True, end='\r')
# # # Get frame id
# # actual_frame_id = int(pcd[13:-8])
# # Get pcd file path
# pcd_file_path = join(scannet_data_path, 'input_pcd_0mean', scene, pcd)