-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathbayesian.py
256 lines (185 loc) · 9.99 KB
/
bayesian.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
import math
import helper
import posenet
import numpy as np
from tensorflow.keras.optimizers import Adam
from scipy.spatial.transform import Rotation
import tensorflow as tf
directory = 'D:\\PycharmProjects\\keras_posenet\\rgb_real\\'
dataset_train_direct = 'dataset_train.txt'
dataset_test_direct = 'dataset_test.txt'
model_arc="posenet"
bayesian=True
if model_arc=="posenet":
f1 = open('D:\\keras-posenet-master\\posenet_angle.txt', 'w')
f2 = open('D:\\keras-posenet-master\\posenet_coord.txt', 'w')
elif model_arc=="lstm-pose":
f1 = open('D:\\keras-posenet-master\\lstmpose_angle.txt', 'w')
f2 = open('D:\\keras-posenet-master\\lstmpose_coord.txt', 'w')
elif model_arc=="poseincepv3":
f1 = open('D:\\keras-posenet-master\\poseincepv3_angle.txt', 'w')
f2 = open('D:\\keras-posenet-master\\poseincepv3_coord.txt', 'w')
elif model_arc=="poselstm_with_2_lstm" :
f1 = open('D:\\keras-posenet-master\\poselstm_with_2_lstm_angle.txt', 'w')
f2 = open('D:\\keras-posenet-master\\poselstm_with_2_lstm_coord.txt', 'w')
def fit_gaussian(pose_quat):
# pose_quat = pose_quat.detach().cpu().numpy()
num_data, _ = pose_quat.shape
# Convert quat to euler
pose_euler = []
for i in range(0, num_data):
pose = pose_quat[i, :3]
quat = pose_quat[i, 3:]
pose_euler.append(np.concatenate((pose, quat)))
# Calculate mean and variance
pose_mean = np.mean(pose_euler, axis=0)
mat_var = np.zeros((7, 7))
for i in range(0, num_data):
pose_diff = pose_euler[i] - pose_mean
mat_var += pose_diff * np.transpose(pose_diff)
mat_var = mat_var / num_data
pose_var = mat_var.diagonal()
return pose_mean, pose_var
if __name__ == "__main__":
# Test model
model = posenet.create_posenet(bayesian)
model.load_weights('custom_trained_weights.h5')
adam = Adam(learning_rate=0.0001, beta_1=0.9, beta_2=0.999, epsilon=1e-08, decay=0.0, clipvalue=2.0)
# model.compile(optimizer=adam, loss={'cls1_fc_pose_xyz': posenet.euc_loss1x, 'cls1_fc_pose_wpqr': posenet.euc_loss1q,
# 'cls2_fc_pose_xyz': posenet.euc_loss2x, 'cls2_fc_pose_wpqr': posenet.euc_loss2q,
# 'cls3_fc_pose_xyz': posenet.euc_loss3x, 'cls3_fc_pose_wpqr': posenet.euc_loss3q})
dataset_train, dataset_test = helper.getKings()
X_test = np.squeeze(np.array(dataset_test.images))
y_test = np.squeeze(np.array(dataset_test.poses))
num_bayesian_test = 3
train_image_names = []
with open(directory + dataset_train_direct) as f:
next(f) # skip the 3 header lines
next(f)
next(f)
for line in f:
fname, p0, p1, p2, p3, p4, p5, p6 = line.split()
train_image_names.append(fname)
def angle_diff_abs(x, y):
x = np.array(x)
y = np.array(y)
error = 180 - abs(abs(x - y) - 180);
return (error)
results = np.zeros((len(dataset_test.images), 6))
for i in range(len(dataset_test.images)):
if bayesian==True:
for k in range(num_bayesian_test):
pos_array = []
ori_array = []
X_test_slice= np.squeeze(X_test[i, :, :, :])
X_test_slice = np.expand_dims(X_test_slice, axis=0)
testPredict = model.predict(X_test_slice)
predicted_x = testPredict[4]
predicted_q = testPredict[5]
predicted_q = np.squeeze(predicted_q)
predicted_q = predicted_q / np.linalg.norm(predicted_q)
predicted_x = np.squeeze(predicted_x)
pos_array.append(predicted_x)
ori_array.append(predicted_q)
pose_q = np.asarray(dataset_test.poses[i][3:7])
pose_x = np.asarray(dataset_test.poses[i][0:3])
pose_q = np.squeeze(pose_q)
pose_q = pose_q / np.linalg.norm(pose_q)
pose_x = np.squeeze(pose_x)
pose_quat = np.concatenate((pos_array, ori_array), 1)
pred_pose, pred_var = fit_gaussian(pose_quat)
pos_var = np.sum(pred_var[:3])
ori_var = np.sum(pred_var[3:])
predicted_x = pred_pose[:3]
predicted_q = pred_pose[3:]
q1 = pose_q / np.linalg.norm(pose_q)
q2 = predicted_q / np.linalg.norm(predicted_q)
d = abs(np.sum(np.multiply(q1, q2)))
theta = 2 * np.arccos(d) * 180 / math.pi
error_distance = np.linalg.norm(pose_x - predicted_x)
predicted_q = np.roll(predicted_q, -1)
pose_q = np.roll(pose_q, -1)
rot_pred = Rotation.from_quat(predicted_q)
rot_real = Rotation.from_quat(pose_q)
rotation_pred = rot_pred.as_euler('zyx', degrees=True)
rotation_real = rot_real.as_euler('zyx', degrees=True)
results[i, :] = [abs((pose_x - predicted_x)[0]), abs((pose_x - predicted_x)[1]),
abs((pose_x - predicted_x)[2]), angle_diff_abs(rotation_pred[0], rotation_real[0]),
angle_diff_abs(rotation_pred[1], rotation_real[1]),
angle_diff_abs(rotation_pred[2], rotation_real[2]),theta,error_distance,pos_var,ori_var]
f2.write(train_image_names[i] + ' ' + str(predicted_x[0]) + ' ' + str(predicted_x[1]) + ' ' + str(
predicted_x[2]) + '\n')
f1.write(train_image_names[i] + ' ' + str(rotation_pred[0]) + ' ' + str(rotation_pred[1]) + ' ' + str(
rotation_pred[2]) + '\n')
print('Image Name: ', train_image_names[i], ' Error X (m): ', abs((pose_x - predicted_x)[0]),
' Error Y (m): ', abs((pose_x - predicted_x)[1]), ' Error Z (m): ',
abs((pose_x - predicted_x)[2]),
' Error Yaw (degrees): ', angle_diff_abs(rotation_pred[0], rotation_real[0]),
' Error Pitch (degrees): '
, angle_diff_abs(rotation_pred[1], rotation_real[1]), ' Error Roll (degrees): ',
angle_diff_abs(rotation_pred[2], rotation_real[2]))
else:
X_test_slice = np.squeeze(X_test[i, :, :, :])
X_test_slice = np.expand_dims(X_test_slice, axis=0)
testPredict = model.predict(X_test_slice)
predicted_x = testPredict[4]
predicted_q = testPredict[5]
pose_q = np.asarray(dataset_test.poses[i][3:7])
pose_x = np.asarray(dataset_test.poses[i][0:3])
pose_q = np.squeeze(pose_q)
pose_q = pose_q / np.linalg.norm(pose_q)
pose_x = np.squeeze(pose_x)
predicted_q = np.squeeze(predicted_q)
predicted_q = predicted_q / np.linalg.norm(predicted_q)
predicted_x = np.squeeze(predicted_x)
q1 = pose_q / np.linalg.norm(pose_q)
q2 = predicted_q / np.linalg.norm(predicted_q)
d = abs(np.sum(np.multiply(q1, q2)))
theta = 2 * np.arccos(d) * 180 / math.pi
error_distance = np.linalg.norm(pose_x - predicted_x)
predicted_q = np.roll(predicted_q, -1)
pose_q = np.roll(pose_q, -1)
rot_pred = Rotation.from_quat(predicted_q)
rot_real = Rotation.from_quat(pose_q)
rotation_pred = rot_pred.as_euler('zyx', degrees=True)
rotation_real = rot_real.as_euler('zyx', degrees=True)
results[i, :] = [abs((pose_x - predicted_x)[0]), abs((pose_x - predicted_x)[1]),
abs((pose_x - predicted_x)[2]), angle_diff_abs(rotation_pred[0], rotation_real[0]),
angle_diff_abs(rotation_pred[1], rotation_real[1]),
angle_diff_abs(rotation_pred[2], rotation_real[2]),theta,error_distance]
f2.write(train_image_names[i] + ' ' + str(predicted_x[0]) + ' ' + str(predicted_x[1]) + ' ' + str(
predicted_x[2]) + '\n')
f1.write(train_image_names[i] + ' ' + str(rotation_pred[0]) + ' ' + str(rotation_pred[1]) + ' ' + str(
rotation_pred[2]) + '\n')
print('Image Name: ', train_image_names[i], ' Error X (m): ', abs((pose_x - predicted_x)[0]),
' Error Y (m): ', abs((pose_x - predicted_x)[1]), ' Error Z (m): ',
abs((pose_x - predicted_x)[2]),
' Error Yaw (degrees): ', angle_diff_abs(rotation_pred[0], rotation_real[0]),
' Error Pitch (degrees): '
, angle_diff_abs(rotation_pred[1], rotation_real[1]), ' Error Roll (degrees): ',
angle_diff_abs(rotation_pred[2], rotation_real[2]))
median_result = np.median(results,axis=0)
if bayesian == False:
print('Median error X ', median_result[0], 'm and ', 'Median error Y ', median_result[1], 'm and ','Median error Z ', median_result[2],
'm and Yaw error ',median_result[3], 'degrees and Pitch error', median_result[4], 'degrees and Roll error', median_result[5], 'degrees.',"angular error",median_result[6]
,"positional error",median_result[7])
else:
print('Median error X ', median_result[0], 'm and ', 'Median error Y ', median_result[1], 'm and ','Median error Z ', median_result[2],
'm and Yaw error ',median_result[3], 'degrees and Pitch error', median_result[4], 'degrees and Roll error', median_result[5], 'degrees.',"angular error",median_result[6]
,"positional error",median_result[7],"positional uncertainity",median_result[8],"angular uncertainity",median_result[9])
f1.close()
f2.close()
f1 = open('D:\\keras-posenet-master\\posenet_angle.txt', 'r')
f3 = open('D:\\keras-posenet-master\\posenet_angle_sorted.txt', 'w')
lines = f1.readlines()
lines_sorted = sorted(lines)
for i in range(len(lines)):
f3.write(lines_sorted[i])
f3.close()
f2 = open('D:\\keras-posenet-master\\posenet_coord.txt', 'r')
f4 = open('D:\\keras-posenet-master\\posenet_coord_sorted.txt', 'w')
lines = f2.readlines()
lines_sorted = sorted(lines)
for i in range(len(lines)):
f4.write(lines_sorted[i])
f4.close()