-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathreduced_3scalesSVM.py
278 lines (256 loc) · 17 KB
/
reduced_3scalesSVM.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
import calculateFeatures
import numpy as np
import laspy as lp
import sys
import send_email
import time
from sklearn import svm
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report, confusion_matrix, accuracy_score
import matplotlib.pyplot as plt
start_read = time.time()
# Get current current time
def get_time():
t = time.localtime()
current_time = time.strftime("%H:%M:%S", t)
return current_time
# FILE PATHS
#LAS files
print(f'Reading LAS files... {get_time()}')
classified_pointCloudPath = '../working/multiscale/classified_sample.las' #change
nonClassified_pointCloudPath = '../working/multiscale/nonClassified_sample.las' #change
#create output txt files
#outputErrorRF = '../results/reduced_error_rf_scale3.txt'
outputErrorSVM = '../results/reduced_error_SVM_scale3.txt'
#create output csv file
#output_path_png = '../results/reduced_gpu_rf_importances.png'
output_path_csv = '../results/reduced_gpu_SVM_scale3.csv'
output_path_las = '../results/reduced_gpu_SVM_scale3.las'
# Read LAS data
# send_email.sendUpdate('Script has begun. Reading LAS files...')
classified_pointCloud = lp.read(classified_pointCloudPath)
classified_points_array = np.vstack((classified_pointCloud.x,
classified_pointCloud.y,
classified_pointCloud.z,
classified_pointCloud['normal z'],
classified_pointCloud.classification,
classified_pointCloud.red,
classified_pointCloud.green,
classified_pointCloud.blue)).transpose()
nonClassified_pointCloud = lp.read(nonClassified_pointCloudPath)
nonClassified_points_array = np.vstack((nonClassified_pointCloud.x,
nonClassified_pointCloud.y,
nonClassified_pointCloud.z,
nonClassified_pointCloud['normal z'],
nonClassified_pointCloud.classification,
nonClassified_pointCloud.red,
nonClassified_pointCloud.green,
nonClassified_pointCloud.blue)).transpose()
# Scales and Radii
#Scale tuples (grid, r): [(0.04, 0.2), (0.08, 0.4), (0.16, 0.8)] #accuracy 90%
#Scale tuples (grid, r): [(0.1, 0.5), (0.2, 1.0), (0.4, 2.0)] #accuracy 93% with 50 trees
grid_sizes = [0.1, 0.2, 0.4]
radii = [0.5, 1.0, 2.0]
print(f'Subsampling classified pc... {get_time()}')
# Subsample the data
classified_subsampled_s1 = calculateFeatures.grid_subsampling_with_color(classified_points_array, grid_sizes[0])
classified_subsampled_s2 = calculateFeatures.grid_subsampling_with_color(classified_points_array, grid_sizes[1])
classified_subsampled_s3 = calculateFeatures.grid_subsampling_with_color(classified_points_array, grid_sizes[2])
print(f'Subsampling nonclassified pc... {get_time()}')
nonClassified_subsampled_s1 = calculateFeatures.grid_subsampling_with_color(nonClassified_points_array, grid_sizes[0])
nonClassified_subsampled_s2 = calculateFeatures.grid_subsampling_with_color(nonClassified_points_array, grid_sizes[1])
nonClassified_subsampled_s3 = calculateFeatures.grid_subsampling_with_color(nonClassified_points_array, grid_sizes[2])
print(f'Calculating geometric features for classified pc... {get_time()}')
# Calculate geometric features for both
classified_features_s1 = calculateFeatures.calculateGeometricFeatures(classified_subsampled_s1, radii[0])
classified_features_s2 = calculateFeatures.calculateGeometricFeatures(classified_subsampled_s2, radii[1])
classified_features_s3 = calculateFeatures.calculateGeometricFeatures(classified_subsampled_s3, radii[2])
print(f'Calculating geometric features for nonclassified pc... {get_time()}')
nonClassified_features_s1 = calculateFeatures.calculateGeometricFeatures(nonClassified_subsampled_s1, radii[0])
nonClassified_features_s2 = calculateFeatures.calculateGeometricFeatures(nonClassified_subsampled_s2, radii[1])
nonClassified_features_s3 = calculateFeatures.calculateGeometricFeatures(nonClassified_subsampled_s3, radii[2])
print(f'Concatenating features... {get_time()}')
# Concatenate the features for classified
classified_Z = np.concatenate([classified_features_s1.get('Z'),classified_features_s2.get('Z'),classified_features_s3.get('Z')])
classified_Z_scaled = np.concatenate([classified_features_s1.get('Z_scaled'),classified_features_s2.get('Z_scaled'),classified_features_s3.get('Z_scaled')])
classified_omnivariance = np.concatenate([classified_features_s1.get('omnivariance'),classified_features_s2.get('omnivariance'),classified_features_s3.get('omnivariance')])
classified_eigenentropy = np.concatenate([classified_features_s1.get('eigenentropy'),classified_features_s2.get('eigenentropy'),classified_features_s3.get('eigenentropy')])
classified_anisotropy = np.concatenate([classified_features_s1.get('anisotropy'),classified_features_s2.get('anisotropy'),classified_features_s3.get('anisotropy')])
classified_linearity = np.concatenate([classified_features_s1.get('linearity'),classified_features_s2.get('linearity'),classified_features_s3.get('linearity')])
classified_planarity = np.concatenate([classified_features_s1.get('planarity'),classified_features_s2.get('planarity'),classified_features_s3.get('planarity')])
classified_curvature = np.concatenate([classified_features_s1.get('curvature'),classified_features_s2.get('curvature'),classified_features_s3.get('curvature')])
classified_sphericity = np.concatenate([classified_features_s1.get('sphericity'),classified_features_s2.get('sphericity'),classified_features_s3.get('sphericity')])
classified_verticality = np.concatenate([classified_features_s1.get('verticality'),classified_features_s2.get('verticality'),classified_features_s3.get('verticality')])
classified_height_range = np.concatenate([classified_features_s1.get('height_range'),classified_features_s2.get('height_range'),classified_features_s3.get('height_range')])
classified_height_avg = np.concatenate([classified_features_s1.get('height_avg'),classified_features_s2.get('height_avg'),classified_features_s3.get('height_avg')])
classified_height_below = np.concatenate([classified_features_s1.get('height_below'),classified_features_s2.get('height_below'),classified_features_s3.get('height_below')])
classified_height_above = np.concatenate([classified_features_s1.get('height_above'),classified_features_s2.get('height_above'),classified_features_s3.get('height_above')])
classified_neighbor_H = np.concatenate([classified_features_s1.get('neighbor_H'),classified_features_s2.get('neighbor_H'),classified_features_s3.get('neighbor_H')])
classified_neighbor_S = np.concatenate([classified_features_s1.get('neighbor_S'),classified_features_s2.get('neighbor_S'),classified_features_s3.get('neighbor_S')])
classified_neighbor_V = np.concatenate([classified_features_s1.get('neighbor_V'),classified_features_s2.get('neighbor_V'),classified_features_s3.get('neighbor_V')])
classified_H_values = np.concatenate([classified_features_s1.get('H'),classified_features_s2.get('H'),classified_features_s3.get('H')])
classified_S_values = np.concatenate([classified_features_s1.get('S'),classified_features_s2.get('S'),classified_features_s3.get('S')])
classified_V_values = np.concatenate([classified_features_s1.get('V'),classified_features_s2.get('V'),classified_features_s3.get('V')])
# Concatenate the features for nonclassified
nonClassified_X = np.concatenate([nonClassified_features_s1.get('X'),nonClassified_features_s2.get('X'),nonClassified_features_s3.get('X')])
nonClassified_Y = np.concatenate([nonClassified_features_s1.get('Y'),nonClassified_features_s2.get('Y'),nonClassified_features_s3.get('Y')])
nonClassified_Z = np.concatenate([nonClassified_features_s1.get('Z'),nonClassified_features_s2.get('Z'),nonClassified_features_s3.get('Z')])
nonClassified_Z_scaled = np.concatenate([nonClassified_features_s1.get('Z_scaled'),nonClassified_features_s2.get('Z_scaled'),nonClassified_features_s3.get('Z_scaled')])
nonClassified_omnivariance = np.concatenate([nonClassified_features_s1.get('omnivariance'),nonClassified_features_s2.get('omnivariance'),nonClassified_features_s3.get('omnivariance')])
nonClassified_eigenentropy = np.concatenate([nonClassified_features_s1.get('eigenentropy'),nonClassified_features_s2.get('eigenentropy'),nonClassified_features_s3.get('eigenentropy')])
nonClassified_anisotropy = np.concatenate([nonClassified_features_s1.get('anisotropy'),nonClassified_features_s2.get('anisotropy'),nonClassified_features_s3.get('anisotropy')])
nonClassified_linearity = np.concatenate([nonClassified_features_s1.get('linearity'),nonClassified_features_s2.get('linearity'),nonClassified_features_s3.get('linearity')])
nonClassified_planarity = np.concatenate([nonClassified_features_s1.get('planarity'),nonClassified_features_s2.get('planarity'),nonClassified_features_s3.get('planarity')])
nonClassified_curvature = np.concatenate([nonClassified_features_s1.get('curvature'),nonClassified_features_s2.get('curvature'),nonClassified_features_s3.get('curvature')])
nonClassified_sphericity = np.concatenate([nonClassified_features_s1.get('sphericity'),nonClassified_features_s2.get('sphericity'),nonClassified_features_s3.get('sphericity')])
nonClassified_verticality = np.concatenate([nonClassified_features_s1.get('verticality'),nonClassified_features_s2.get('verticality'),nonClassified_features_s3.get('verticality')])
nonClassified_height_range = np.concatenate([nonClassified_features_s1.get('height_range'),nonClassified_features_s2.get('height_range'),nonClassified_features_s3.get('height_range')])
nonClassified_height_avg = np.concatenate([nonClassified_features_s1.get('height_avg'),nonClassified_features_s2.get('height_avg'),nonClassified_features_s3.get('height_avg')])
nonClassified_height_below = np.concatenate([nonClassified_features_s1.get('height_below'),nonClassified_features_s2.get('height_below'),nonClassified_features_s3.get('height_below')])
nonClassified_height_above = np.concatenate([nonClassified_features_s1.get('height_above'),nonClassified_features_s2.get('height_above'),nonClassified_features_s3.get('height_above')])
nonClassified_neighbor_H = np.concatenate([nonClassified_features_s1.get('neighbor_H'),nonClassified_features_s2.get('neighbor_H'),nonClassified_features_s3.get('neighbor_H')])
nonClassified_neighbor_S = np.concatenate([nonClassified_features_s1.get('neighbor_S'),nonClassified_features_s2.get('neighbor_S'),nonClassified_features_s3.get('neighbor_S')])
nonClassified_neighbor_V = np.concatenate([nonClassified_features_s1.get('neighbor_V'),nonClassified_features_s2.get('neighbor_V'),nonClassified_features_s3.get('neighbor_V')])
nonClassified_H_values = np.concatenate([nonClassified_features_s1.get('H'),nonClassified_features_s2.get('H'),nonClassified_features_s3.get('H')])
nonClassified_S_values = np.concatenate([nonClassified_features_s1.get('S'),nonClassified_features_s2.get('S'),nonClassified_features_s3.get('S')])
nonClassified_V_values = np.concatenate([nonClassified_features_s1.get('V'),nonClassified_features_s2.get('V'),nonClassified_features_s3.get('V')])
#Stack features for classification
print(f'Stacking features... {get_time()}')
classified_features = np.vstack((
classified_omnivariance,
classified_eigenentropy,
classified_anisotropy,
classified_linearity,
classified_planarity,
classified_curvature,
classified_sphericity,
classified_verticality,
classified_Z_scaled,
classified_height_range,
classified_height_avg,
classified_height_below,
classified_height_above,
classified_neighbor_H,
classified_neighbor_S,
classified_neighbor_V,
classified_H_values,
classified_S_values,
classified_V_values
)).transpose()
nonClassified_features = np.vstack((
nonClassified_omnivariance,
nonClassified_eigenentropy,
nonClassified_anisotropy,
nonClassified_linearity,
nonClassified_planarity,
nonClassified_curvature,
nonClassified_sphericity,
nonClassified_verticality,
nonClassified_Z_scaled,
nonClassified_height_range,
nonClassified_height_avg,
nonClassified_height_below,
nonClassified_height_above,
nonClassified_neighbor_H,
nonClassified_neighbor_S,
nonClassified_neighbor_V,
nonClassified_H_values,
nonClassified_S_values,
nonClassified_V_values
)).transpose()
# Labels
labels = np.concatenate([classified_features_s1.get('classification'),classified_features_s2.get('classification'),classified_features_s3.get('classification')])
#-------#
send_email.sendUpdate('Classification has begun')
# Train a classifier
X_train, X_test, y_train, y_test = train_test_split(classified_features, labels, test_size=0.2, random_state=42)
# Machine learning models
#rf_model = RandomForestClassifier()
svm_model = svm.SVC()
# Train the models
print(f'Training models... {get_time()}')
#rf_model.fit(X_train, y_train)
svm_model.fit(X_train, y_train)
# Evaluate model
send_email.sendUpdate('Training done. Evaluating models...')
print(f'Evaluating models... {get_time()}')
#y_pred_rf = rf_model.predict(X_test)
y_pred_svm = svm_model.predict(X_test)
#Predict the non-classified area
# RF model report
# print(f'Writing RF classification performance to file... {get_time()}')
# report_RF = classification_report(y_test, y_pred_rf)
# matrix_RF = confusion_matrix(y_test, y_pred_rf)
# accuracy_RF = accuracy_score(y_test, y_pred_rf)
# importances = rf_model.feature_importances_
features = ['omnivariance',
'eigenentropy',
'anisotropy',
"linearity",
"planarity",
"curvature",
"sphericity",
"verticality",
"Z values",
"height_range",
"height_avg",
"height_below",
"height_above",
"neighbor_H",
"neighbor_S",
"neighbor_V",
"H_values",
"S_values",
"V_values"]
#Get accuracy results and write to file
# with open(outputErrorRF, 'w') as f:
# f.write('Classification Report for Random Forests:\n')
# f.write(report_RF)
# f.write('\nConfusion Matrix:\n')
# f.write(str(matrix_RF))
# f.write(f'\nAccuracy: {accuracy_RF * 100:.2f}%')
# f.write('\nFeature ranking:\n')
# for f_index in range(len(features)):
# f.write(f"{features[f_index]}: {importances[f_index]}\n")
# SVM model report
print(f'Writing SVM classification performance to file... {get_time()}')
report_svm = classification_report(y_test, y_pred_svm)
matrix_svm = confusion_matrix(y_test, y_pred_svm)
accuracy_svm = accuracy_score(y_test, y_pred_svm)
# Write the results to file
with open(outputErrorSVM, 'w') as f:
f.write('Classification Report for SVM:\n')
f.write(report_svm)
f.write('\nConfusion Matrix:\n')
f.write(str(matrix_svm))
f.write(f'\nAccuracy: {accuracy_svm * 100:.2f}%')
print(f'Predicting non-classified pc... {get_time()}')
send_email.sendUpdate('Predicting on unseen data. Model performance written to file.')
#predictions_RF = rf_model.predict(nonClassified_features)
predictions_SVM = svm_model.predict(nonClassified_features)
result_output_array= np.vstack((nonClassified_X,
nonClassified_Y,
nonClassified_Z,
predictions_SVM,
nonClassified_verticality #place holder second ML values
)).T
print(f'Saving CSV file... {get_time()}')
#write to txt file in case las write didnt work
#np.savetxt(output_path_csv,result_output_array, delimiter=',',header='X,Y,Z,RF,GBT',comments='')
try:
print(f'Saving classified points as LAS... {get_time()}')
calculateFeatures.saveNP_as_LAS(result_output_array, # Array with X,Y,Z values
nonClassified_pointCloud, # Reference pc with headers
output_path_las, # output path
nonClassified_verticality, #place holder second ML values
predictions_SVM) # SVM values
except Exception as e:
print(e)
send_email.sendNotification('Error in saving classified points as LAS')
eval_time = round((time.time() - start_read)/3600,2)
print(f"Evaluated time: {eval_time} hours")
# Add mailme to CLI and get an email notification sent when scipt is done
if len(sys.argv) > 1:
if sys.argv[1]=='mailme':
send_email.sendNotification(f"""Process finished. Classification of data is done.
\nThe whole process elapsed {eval_time} hours""")