-
Notifications
You must be signed in to change notification settings - Fork 0
/
dataIO.py
362 lines (298 loc) · 13.2 KB
/
dataIO.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
"""
dataIO for 'original' dataset, support both tsinghua dataset and HMD dataset
"""
import glob as glob
import os
import numpy as np
from matplotlib import pyplot as plt
# from sklearn.gaussian_process import GaussianProcessRegressor
# from sklearn.gaussian_process.kernels import RBF, WhiteKernel, ConstantKernel as C
np.random.seed(1)
import pdb
import _pickle as pickle
from config import cfg
import math
def clip_xyz(x):
"""clip xyz to [-1,1]"""
# for ii in range(len(x.keys())):
# print(x.keys())
for ii in x.keys():
x[ii][b'x'][x[ii][b'x']<-1]=-1
x[ii][b'x'][x[ii][b'x']>1]=1
x[ii][b'y'][x[ii][b'y']<-1]=-1
x[ii][b'y'][x[ii][b'y']>1]=1
x[ii][b'z'][x[ii][b'z']<-1]=-1
x[ii][b'z'][x[ii][b'z']>1]=1
return x
def clip_xyz0(x):
"""clip xyz to [-1,1]"""
# for ii in range(len(x.keys())):
# print(x.keys())
for ii in x.keys():
x[ii]['x'][x[ii]['x']<-1]=-1
x[ii]['x'][x[ii]['x']>1]=1
x[ii]['y'][x[ii]['y']<-1]=-1
x[ii]['y'][x[ii]['y']>1]=1
x[ii]['z'][x[ii]['z']<-1]=-1
x[ii]['z'][x[ii]['z']>1]=1
return x
def quaternion2euler2(q0,q1,q2,q3):
q0 = np.float(q0)
q1 = np.float(q1)
q2 = np.float(q2)
q3 = np.float(q3)
"""convert quaternion tuple to euler angles"""
roll = np.arctan2(2*(q0*q1+q2*q3),(1-2*(q1**2+q2**2)))
# confine to [-1,1] to avoid nan from arcsin
sintemp = min(1,2*(q0*q2-q3*q1))
sintemp = max(-1,sintemp)
pitch = np.arcsin(sintemp)
yaw = np.arctan2(2*(q0*q3+q1*q2),(1-2*(q2**2+q3**2)))
return roll,pitch,yaw
def quaternion2euler3(qx,qy,qz,qw):
"""from tsinghua paper"""
qx = np.float(qx)
qy = np.float(qy)
qz = np.float(qz)
qw = np.float(qw)
"""convert quaternion tuple to euler angles"""
x = 2*qx*qz+2*qy*qw
y = 2*qy*qz-2*qx*qw
z = 1-2*(qx**2)-2*(qy**2)
return x,y,z
# def quaternion2euler4(yaw,pitch):
# # from kanishk's report
# x = np.cos(yaw)*np.cos(pitch)
# y = np.sin(yaw)*np.cos(pitch)ßß
# z = np.sin(pitch)
# return x,y,z
def quaternion2euler4(yaw,pitch):
# from CUB360 paper
z = np.cos(yaw)*np.cos(pitch)
x = np.sin(yaw)*np.cos(pitch)
y = np.sin(pitch)
return x,y,z
def xyz2thetaphi_thu(x,y,z): #same as in dataIO_tsinghua
# convert x,y,z to theta and phi
# theta = arctan(y/x)
theta = np.arctan2(y,x) #return [-pi,pi]
phi = np.arctan2(np.sqrt(x**2+y**2),z)
return theta, phi
def xyz2thetaphi(x,y,z):
# convert x,y,z to original theta and phi(input for lat_long2xyz())
# i.e. theta [-pi,pi], phi [0,pi]
theta = np.mod(np.arctan2(y,x),2*np.pi)-np.pi
phi = np.mod(np.arctan2(z,np.sqrt(x**2+y**2))+np.pi/2, np.pi)
return theta, phi
def get_anglelist_from_file_HDM(temp):
"""for HMD dataset"""
"""get euler angle list from the file lines"""
roll_list = []
pitch_list = []
yaw_list = []
for ii in range(len(temp)):
[q0,q1,q2,q3] = np.array(temp[ii].split(' '))[-4:]
roll,pitch,yaw = quaternion2euler2(q0,q1,q2,q3)
roll_list.append(roll)
pitch_list.append(pitch)
yaw_list.append(yaw)
return roll_list,pitch_list,yaw_list
def get_anglelist_from_file_tsinghua(temp):
"""for Tsinghua dataset"""
"""get euler angle list from the file lines"""
roll_list = []
pitch_list = []
yaw_list = []
# HmdPosition_x_list = []
# HmdPosition_y_list = []
# HmdPosition_z_list = []
for ii in range(1,len(temp)):
[qx,qy,qz,qw] = np.array(temp[ii].split(','))[2:6]
#!!!order not one to one! https://en.wikipedia.org/wiki/Conversion_between_quaternions_and_Euler_angles
q0,q1,q2,q3 = qw,qx,qy,qz
roll,pitch,yaw = quaternion2euler2(q0,q1,q2,q3)
roll_list.append(roll)
pitch_list.append(pitch)
yaw_list.append(yaw)
# [x,y,z] = np.array(temp[ii].split(','))[6:]
# HmdPosition_x_list.append(x)
# HmdPosition_y_list.append(y)
# HmdPosition_z_list.append(z)
return roll_list,pitch_list,yaw_list
def get_xyz_from_file_tsinghua(temp):
"""for Tsinghua dataset"""
x_list = []
y_list = []
z_list = []
for ii in range(1,len(temp)):
[qx,qy,qz,qw] = np.array(temp[ii].split(','))[2:6]
x,y,z = quaternion2euler3(qx,qy,qz,qw)
x_list.append(x)
y_list.append(y)
z_list.append(z)
return x_list,y_list,z_list
def raw_2_sincos(angle_array):
"""convert raw angles (yaw,pitch,roll) into (sin, cos) pairs"""
cos_val = np.cos(angle_array)
sin_val = np.sin(angle_array)
return (cos_val,sin_val)
def _get_length(video_ind):
# get the rough length, cut the tail
length = 1e8
for user_ind in range(1,user_num+1):
filepath = directory+str(user_ind)+'/video_'+str(video_ind)+'.csv'
temp = open(filepath,'r').readlines()
length = min(len(temp)-1,length)
# print('video_ind', video_ind, 'length ',length)
return int(length)
def _get_time_duration(watch_time_start,watch_time_end):
hr,mt,sd = np.float32(watch_time_start.split(':'))
hr1,mt1,sd1 = np.float32(watch_time_end.split(':'))
duration = (hr1-hr)*3600+(mt1-mt)*60+(sd1-sd)
return duration
def discretize360(x_list,y_list,z_list,bin_size=30):
"""quantize the sphere into one-hot vector, each block is (bin_size,bin_size)"""
n = len(x_list)
one_hot_code_matrix = np.zeros((n,180/bin_size,360/bin_size))
for i in range(n):
theta,phi = xyz2thetaphi(x_list[i],y_list[i],z_list[i])
theta = theta/np.pi*180
phi = phi/np.pi*180
col = math.floor(theta/bin_size)
row = math.floor(phi/bin_size)
if phi == 180:
row = 5
if theta == 180:
col = 5
#print(theta)
#print(phi)
#one_hot_code_matrix[i,0] = int(row)
#one_hot_code_matrix[i,1] = int(col+6)
#one_hot_code = np.zeros((6,12))
one_hot_code_matrix[i,int(row),int(col+6)] = 1
#one_hot_code_matrix[i,:] = one_hot_code.flatten()
return one_hot_code_matrix
def reshape_one_user(one_hot_code_matrix):
"""reshape the data of one user for one video to n-by-30-by-6-by-12 matrix"""
#n is secs
#input size is numframes-by-6-by-12
#output size is numsec-by-30-by-6-by-12
nframes = one_hot_code_matrix.shape[0]
nsecs = nframes//30
new_representation = np.zeros((nsecs,30,6,12)) #ignore the remainder here
for i in range(nsecs):
new_representation[i,:,:,:] = one_hot_code_matrix[i*30:(i+1)*30,:,:]
return new_representation
def reshape_multi_users(one_hot_code_matrix_of_multiusers):
"""reshape the data of multiple users for one video to one single n-by-30-by-6-by-12 matrix"""
#n is secs
#input size is numusers-by-numframes-by-6-by-12
#output size is numsec-by-30-by-6-by-12
nusers = one_hot_code_matrix_of_multiusers.shape[0]
nframes = one_hot_code_matrix_of_multiusers.shape[1]
nsecs = nframes//30
new_representation = np.zeros((nsecs,30,6,12)) #ignore the remainder
for i in range(nsecs):
temp = np.sum(one_hot_code_matrix_of_multiusers[:,i*30:(i+1)*30,:,:],0)
new_representation[i,:,:,:] = temp
return new_representation
def gen_heatmap_like_matrix(one_hot_code_matrix):
"""generate a heatmap-like matrix for one user"""
#input size is numframes-by-6-by-12
#output size is 6-by-12
heatmap = np.zeros((6,12))
heatmap = np.sum(one_hot_code_matrix,0)
return heatmap
def gen_heatmap_like_matrix_for_multiusers(one_hot_code_matrix_of_multiusers):
"""generate a heatmap-like matrix for one user"""
#input size is numusers-by-numframes-by-6-by-12
#output size is 6-by-12
heatmap = np.zeros((6,12))
heatmap = np.sum(np.sum(one_hot_code_matrix_of_multiusers,0),0)
return heatmap
if __name__ == '__main__':
# Tsinghua Dataset
experiment = 1
experimentpath = 'Experiment_'+str(experiment)+'/'
directory = os.path.join('./360video/tsinghua/Formated_Data/',experimentpath)
user_num = 48
# video info
videoMetaFile = os.path.join('./360video/tsinghua/Formated_Data/',experimentpath,'videoMeta.csv')
videoMeta = open(videoMetaFile,'r').readlines()
all_video_data = {} #from all videos of one experiment
all_video_data_pair = {}
all_video_data_xyz = {}
duration_dic = {}
all_video_data_phi_theta = {}
for video_ind in range(9):
video_info = videoMeta[video_ind+1].split(',')
fps = int(video_info[3])
vid_duration = int(video_info[2].split(':')[0])*60+int(video_info[2].split(':')[1])
print('video ',video_ind,' has duration of: ', vid_duration,' seconds')
duration_dic[video_ind] = {}
shortest_len = _get_length(video_ind)
if cfg.use_yaw_pitch_roll:
all_video_data[video_ind] = {}
all_video_data[video_ind]['raw_roll'] = np.zeros((user_num,shortest_len))
all_video_data[video_ind]['raw_pitch'] = np.zeros((user_num,shortest_len))
all_video_data[video_ind]['raw_yaw'] = np.zeros((user_num,shortest_len))
if cfg.use_xyz:
all_video_data_xyz[video_ind] = {}
all_video_data_xyz[video_ind]['x'] = np.zeros((user_num,shortest_len))
all_video_data_xyz[video_ind]['y'] = np.zeros((user_num,shortest_len))
all_video_data_xyz[video_ind]['z'] = np.zeros((user_num,shortest_len))
if cfg.use_phi_theta:
all_video_data_phi_theta[video_ind] = {}
all_video_data_phi_theta[video_ind]['phi'] = np.zeros((user_num,shortest_len))
all_video_data_phi_theta[video_ind]['theta'] = np.zeros((user_num,shortest_len))
for user_ind in range(1,user_num+1):
filepath = directory+str(user_ind)+'/video_'+str(video_ind)+'.csv'
temp = open(filepath,'r').readlines()
# watch_time_start = temp[1].split(',')[0].split(' ')[1]
# watch_time_end = temp[-1].split(',')[0].split(' ')[1]
# duration_dic[video_ind][user_ind] = _get_time_duration(watch_time_start,watch_time_end)
# print('user ',user_ind,' watched: ', duration_dic[video_ind][user_ind])
if cfg.use_xyz:
x_list,y_list,z_list = get_xyz_from_file_tsinghua(temp)
all_video_data_xyz[video_ind]['x'][user_ind-1,:] = x_list[:shortest_len]
all_video_data_xyz[video_ind]['y'][user_ind-1,:] = y_list[:shortest_len]
all_video_data_xyz[video_ind]['z'][user_ind-1,:] = z_list[:shortest_len]
elif cfg.use_yaw_pitch_roll:
roll_list,pitch_list,yaw_list = get_anglelist_from_file_tsinghua(temp)
all_video_data[video_ind]['raw_roll'][user_ind-1,:] = roll_list[:shortest_len]
all_video_data[video_ind]['raw_pitch'][user_ind-1,:] = pitch_list[:shortest_len]
all_video_data[video_ind]['raw_yaw'][user_ind-1,:] = yaw_list[:shortest_len]
elif cfg.use_phi_theta:
x_list,y_list,z_list = get_xyz_from_file_tsinghua(temp)
theta_list, phi_list = xyz2thetaphi(np.array(x_list),np.array(y_list),np.array(z_list))
all_video_data_phi_theta[video_ind]['phi'][user_ind-1,:] = phi_list[:shortest_len]
all_video_data_phi_theta[video_ind]['theta'][user_ind-1,:] = theta_list[:shortest_len]
pdb.set_trace()
# TODO: test whether the same with tsinghua paper
x_list2,y_list2,z_list2 = quaternion2euler4(yaw_list[:shortest_len],pitch_list[:shortest_len])
# print('video_ind ',video_ind,'len(yaw_list) ',len(yaw_list))
###lengths are different, why?
# print user_ind,len(yaw_list)
# print(temp[1].split(',')[0].split(' ')[1], temp[-1].split(',')[0].split(' ')[1])
if cfg.use_cos_sin:
all_video_data_pair[video_ind] = {}
# (cos,sin) pair
all_video_data_pair[video_ind]['cos_yaw'] = np.zeros((user_num,shortest_len))
all_video_data_pair[video_ind]['sin_yaw'] = np.zeros((user_num,shortest_len))
all_video_data_pair[video_ind]['cos_pitch'] = np.zeros((user_num,shortest_len))
all_video_data_pair[video_ind]['sin_pitch'] = np.zeros((user_num,shortest_len))
cos_yaw,sin_yaw = raw_2_sincos(all_video_data[video_ind]['raw_yaw'])
cos_pitch,sin_pitch = raw_2_sincos(all_video_data[video_ind]['raw_pitch'])
# cos_roll,sin_roll = raw_2_sincos(all_video_data[video_ind]['raw_roll'])
all_video_data_pair[video_ind]['cos_yaw'] = cos_yaw
all_video_data_pair[video_ind]['sin_yaw'] = sin_yaw
all_video_data_pair[video_ind]['cos_pitch'] = cos_pitch
all_video_data_pair[video_ind]['sin_pitch'] = sin_pitch
if cfg.use_yaw_pitch_roll:
pickle.dump(all_video_data,open('./data/exp_'+str(experiment)+'_raw.p','wb'))
if cfg.use_cos_sin:
pickle.dump(all_video_data_pair,open('./data/exp_'+str(experiment)+'_raw_pair.p','wb'))
if cfg.use_xyz:
pickle.dump(all_video_data_xyz,open('./data/exp_'+str(experiment)+'_xyz.p','wb'))
if cfg.use_phi_theta:
pickle.dump(all_video_data_phi_theta,open('./data/exp_'+str(experiment)+'_phi_theta.p','wb'))