-
Notifications
You must be signed in to change notification settings - Fork 1
/
GaitLab2Go.py
285 lines (244 loc) · 11.4 KB
/
GaitLab2Go.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
import pickle
import numpy as np
import time
import matplotlib.pyplot as plt
import os
import pandas as pd
import glob
from scipy import interpolate
import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
from tensorflow.keras.layers import Flatten
from tensorflow.keras.layers import Dropout
from tensorflow.keras.layers import Conv1D
from tensorflow.keras.layers import MaxPooling1D
from tensorflow.keras.layers import BatchNormalization
from tensorflow.keras.layers import AveragePooling1D
class data_processing(object):
def __init__(self):
self.enviroment='GaitLab2Go'
def find_files(self,path='.',ext='.xlsx'):
serach_term=path +'/**/*'+ ext
X=np.array([])
i=0
for file_name in glob.iglob(serach_term, recursive=True):
Y=np.array([file_name])
X=np.append(X,Y)
i=+1
return X
def randomize(self,LL):
A=np.array(list(range(LL)))
np.random.shuffle(A)
return A
def string2int(self,data,colum_index=1):
unique_identity=np.unique(data[:,colum_index])
self.identity=unique_identity
for i in range(unique_identity.shape[0]):
data[:,colum_index]=np.where(data[:,colum_index]==unique_identity[i],int(i),data[:,colum_index])
return data
def train_test_split(self,data,parameters,subject_index=-2,subjects_wise_random=True,
validation=False,output_parameters='all',
prediction_index=-1,test_percentage=0.25,
normalization=True):
if output_parameters=='all':
out_colum_idx = np.delete(np.arange(len(parameters)),[subject_index,prediction_index])
#out_colum_idx= np.delete(out_colum_idx,prediction_index-1)
else:
out_colum_idx = np.where(np.char.find(list(parameters),output_parameters)==0)[0]
n_x,n_y= data[:,out_colum_idx],data[:,prediction_index]
if subjects_wise_random:
subjects=np.unique(data[:,subject_index])
test_subjects=int(round(subjects.shape[0]*test_percentage))
test_inds=self.randomize(subjects.shape[0])[0:test_subjects]
test_subjects=subjects[test_inds]
train_inds=np.delete(np.arange(subjects.shape[0]),test_inds)
train_subjects=subjects[train_inds]
self.train_subjects=train_subjects
self.test_subjects=test_subjects
inds={'train':np.array([]),'test':np.array([]),'validation':np.array([])}
for subject in range(train_subjects.shape[0]):
inds['train']=np.append(inds['train'],np.where(data[:,subject_index]==train_subjects[subject])[0])
for subject in range(test_subjects.shape[0]):
inds['test']=np.append(inds['test'],np.where(data[:,subject_index]==test_subjects[subject])[0])
if validation:
pass
x_train,y_train=n_x[inds['train'].astype('int64'),:],n_y[inds['train'].astype('int64')]
x_test,y_test=n_x[inds['test'].astype('int64'),:],n_y[inds['test'].astype('int64')]
train_idexs=self.randomize(y_train.shape[0])
test_idexs=self.randomize(y_test.shape[0])
x_train,y_train,x_test,y_test=x_train[train_idexs],y_train[train_idexs],x_test[test_idexs],y_test[test_idexs]
if validation:
pass
else:
test_num_idex=int(round(n_y.shape[0]*test_percentage))
train_num_idex=int(n_y.shape[0]-test_num_idex)
if validation:
pass
idexs=self.randomize(n_y.shape[0])
test_idex,train_idex=idexs[0:test_num_idex],idexs[test_num_idex+1:train_num_idex]
x_train,y_train=n_x[train_idex,:],n_y[train_idex]
x_test,y_test=n_x[test_idex,:],n_y[test_idex]
self.x_train,self.y_train,self.x_test,self.y_test=x_train,y_train,x_test,y_test
self.output_parameters,self.subjects_wise_random=output_parameters,subjects_wise_random
if normalization:
self.normalize()
def normalize(self):
from sklearn import preprocessing
self.scaler = preprocessing.StandardScaler().fit(self.x_train)
self.x_train=self.scaler.transform(self.x_train)
self.x_test=self.scaler.transform(self.x_test)
#self.valid[0]=self.scaler.transform(self.valid[0])
#self.x_mean = np.mean(self.x_train,axis=0)
#self.x_std = np.std(self.x_train,axis=0)
#self.x_train=(self.x_train-self.x_mean)/self.x_std
#self.y_train=(self.x_test-self.x_mean)/self.x_std
def ForwardFeed(self):
import ForwardFeed as FF
self.FF=FF
def Normalized_gait(self,A):
x=np.arange(A.shape[-1])
A=np.where(np.isnan(A)==1,0,A)
Y= interpolate.interp1d(x,A, kind='cubic')(np.linspace(x.min(), x.max(), 101))
return Y
def convert_data(self,data):
allpara=list(data.columns)
parameters=allpara[0:-2]
norm_data={}
for para in parameters:
norm_data[f"{para}"]=[]
for j in range(data.shape[0]):
A=data.iloc[j,0:-2].to_numpy()
normdata=np.zeros([A.shape[0],A[0].shape[0]],dtype='float64')
for i in range (A.shape[0]):
if A[i].shape[0]>A[i].shape[1]:
normdata[i,:]=A[i].reshape(A[i].shape[0])
elif A[i].shape[0]<A[i].shape[1]:
normdata[i,:]=A[i].reshape(A[i].shape[1])
new=self.Normalized_gait(normdata)
for para in range(len(parameters)):
norm_data[f"{parameters[para]}"].append(new[para])
self.Ndata=norm_data
self.Ndata["Subjects"]=data[allpara[-2]].values
self.Ndata["Surface"]=data[allpara[-1]].values
def surface_data(self):
key=list(self.Ndata.keys())
mdata=pd.read_pickle('mean_std_surface.plk')
surface=mdata.to_dict()
subject={}
for i in range(len(self.identity)):
indexs=np.where(self.Ndata["Surface"]==i)[0]
subject[f"{self.identity[i]}"]=np.array(self.Ndata["Subjects"])[indexs]
for j in range(len(key)-2):
surface[f"{self.identity[i]}"][f"{key[j]}"]["data"]=np.array(self.Ndata[f"{key[j]}"])[indexs]
self.Sdata=surface
self.subjects=subject
def feature_extraction(self):
import features as fs
mdata=pd.read_pickle('mean_std_surface.plk')
para=list(self.Ndata.keys())
self.features={}
for i in range(len(para)-int(2)):
outdata=pd.DataFrame(self.Ndata[para[i]])
self.features[f'{para[i]}']=fs.feature(outdata,mdata,para[i])
self.features[para[-2]]=self.Ndata[para[-2]]
self.features[para[-1]]=self.Ndata[para[-1]]
def all_features(self):
para=list(self.features.keys())
self.f={}
for i in range(len(para)-int(2)):
fetur=list(self.features[para[i]].keys())
for j in range(len(fetur)):
self.f[f'{para[i]}_{fetur[j]}']=self.features[para[i]][fetur[j]]
self.f[para[-2]]=self.features[para[-2]]
self.f[para[-1]]=self.features[para[-1]]
self.computed_features=fetur
self.sensors=para
def ANN(self,hid_layers,model,output,input_shape,activation_hid='relu'):
ann = tf.keras.models.Sequential()
ann.add(tf.keras.Input(shape=input_shape))
for l in hid_layers:
ann.add(tf.keras.layers.Dense(units=l,activation=activation_hid))
if model=='classification':
ann.add(tf.keras.layers.Dense(units=output,activation='softmax'))
ann.compile(optimizer='adam',loss='categorical_crossentropy',metrics=['accuracy'])
return ann
elif model=='regression':
ann.add(tf.keras.layers.Dense(units=output))
ann.compile(optimizer='adam',loss='mean_squared_error')
return ann
def CNN_test(self,input_shape,output_shape):
model = tf.keras.models.Sequential()
model.add(Conv1D(filters=64, kernel_size=3, activation='relu', input_shape=input_shape))
model.add(Conv1D(filters=64, kernel_size=3, activation='relu'))
model.add(Dropout(0.5))
model.add(MaxPooling1D(pool_size=2))
model.add(Flatten())
model.add(Dense(100, activation='relu'))
model.add(Dense(output_shape, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
return model
def CNN_Hu(self,input_shape,output_shape):
model = tf.keras.models.Sequential()
model.add(Conv1D(filters=100, kernel_size=3, activation='relu', input_shape=input_shape))
model.add(BatchNormalization())
model.add(Conv1D(filters=100, kernel_size=3, activation='relu'))
model.add(BatchNormalization())
model.add(Dropout(0.2))
model.add(Conv1D(filters=100, kernel_size=3, activation='relu'))
model.add(BatchNormalization())
model.add(Conv1D(filters=50, kernel_size=3, activation='relu'))
model.add(BatchNormalization())
model.add(MaxPooling1D(pool_size=3))
model.add(Dropout(0.2))
model.add(Conv1D(filters=50, kernel_size=3, activation='relu'))
model.add(BatchNormalization())
model.add(Conv1D(filters=50, kernel_size=3, activation='relu'))
model.add(BatchNormalization())
model.add(Dropout(0.2))
model.add(AveragePooling1D())
model.add(Flatten())
#model.add(Dense(100, activation='relu'))
model.add(Dense(output_shape, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
return model
def CNN1D(self,filters,kernel_size,input_shape,pool_size,model,hid_layers,activation_hid,output):
cnn1d = tf.keras.models.Sequential()
for i in range(len(filters)):
if i==0:
cnn1d.add(tf.keras.layers.Conv1D(filters[i],kernel_size[i],activation=activation_hid,input_shape=input_shape))
cnn1d.add(tf.keras.layers.MaxPooling1D(pool_size[i]))
else:
cnn1d.add(tf.keras.layers.Conv1D(filters[i],kernel_size[i],activation=activation_hid))
cnn1d.add(tf.keras.layers.MaxPooling1D(pool_size[i]))
cnn1d.add(tf.keras.layers.Flatten())
for l in hid_layers:
cnn1d.add(tf.keras.layers.Dense(units=l,activation=activation_hid))
if model=='classification':
cnn1d.add(tf.keras.layers.Dense(units=output,activation='softmax'))
cnn1d.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
return cnn1d
elif model=='regression':
cnn1d.add(tf.keras.layers.Dense(units=output))
cnn1d.compile(optimizer='adam',loss='mean_squared_error')
return cnn1d
def one_hot(self,y):
uniq=np.unique(y)
y_hot=np.zeros([y.shape[0],uniq.shape[0]])
for i in range(len(uniq)):
index=np.where(y==uniq[i])[0]
y_hot[index,i]=1
self.surface_name=uniq
return y_hot
def one_hot_y(self,y):
uniq=self.surface_name
y_hot=np.zeros([y.shape[0],uniq.shape[0]])
for i in range(len(uniq)):
index=np.where(y==uniq[i])[0]
y_hot[index,i]=1
self.surface_name=uniq
return y_hot
if __name__ == "__main__":
# WRITE CODE HERE
# Instantiate, train, and evaluate your classifiers in the space below
pass