-
Notifications
You must be signed in to change notification settings - Fork 0
/
classify_LSTM_correct.py
265 lines (209 loc) · 8.41 KB
/
classify_LSTM_correct.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
# LSTM for phone recognition with input a-b lip
import numpy
import matplotlib.pyplot as plt
import pandas
import math
from keras.models import Sequential
#from keras.models import Graph
from keras.layers import Dense
from keras.layers import LSTM
#from keras.layers import TimeDistributedDense
from keras.layers import TimeDistributed
from keras.layers import Activation
from keras.layers import Dropout
from keras.layers import Input
from keras.models import Model
from sklearn.preprocessing import MinMaxScaler
from sklearn.metrics import mean_squared_error
from keras.models import load_model
import h5py
from sklearn.preprocessing import LabelEncoder
from keras.utils import np_utils
import scipy.io as spio
#for nb_Cell in [10,15]:
workdir = '/media/liliu/Sony_LiLIU/linux_liliu_beifen/Lip_recognition/Hand_position_experiment_88/'
nb_Cell = 50
#print('nb_Cell = %d' % nb_Cell)
nb_epoch= 10
batchsize = 1
#nb_sbj = 10
#nb_w = 50
#nb_file = nb_sbj*nb_w
nb_file = 88
nb_file_train = int(0.7*nb_file)
nb_file_test = int(0.3*nb_file)
drop = 0.5
dim_input = 2
normalization_coeff = 1
# fix random seed for reproducibility
#numpy.random.seed(7)
#maxVal_IU,maxVal_MP,maxVal_SP,maxVal_GT,maxVal_FX = 6,5,5,5,8
maxVal_Phone = 5 #6
def to_categorical(seq_data,maxVal):
category_data = numpy.zeros((len(seq_data),maxVal))
for i in range(len(seq_data)):
if seq_data[i] >= maxVal:
category_data[i,maxVal-1] = 1.0
else:
category_data[i,int(seq_data[i])-1]=1.0
return category_data
def find_max(n):
max_len = -1
for i in range(n):
filename = workdir +'data_combine_code_1D/data_'+str(i+1);
data_in = pandas.read_csv(filename, usecols=[0], engine='python',header=None, sep=r"\s+")
len_data = len(data_in)
if max_len < len_data:
max_len = len_data
return max_len
max_len = find_max(nb_file)
def padding_zero(data_):
len_data_ = len(data_)
if len_data_ < max_len:
add_zeros = numpy.zeros(max_len-len_data_,dtype=numpy.int)
data_padding = numpy.concatenate((data_,add_zeros),axis = 0)
else:
data_padding = data_
return data_padding
def padding_zero_one(data_):
len_data_ = len(data_)
if len_data_ < max_len:
add_ones = numpy.ones(max_len-len_data_,dtype=numpy.int)
data_padding = numpy.concatenate((data_,add_ones),axis = 0)
else:
data_padding = data_
return data_padding
# def TimeDistributedDense(x, w, b=None, dropout=None,
# input_dim=None, output_dim=None, timesteps=None):
# '''Apply y.w + b for every temporal slice y of x.
# '''
# if not input_dim:
# # won't work with TensorFlow
# input_dim = K.shape(x)[2]
# if not timesteps:
# # won't work with TensorFlow
# timesteps = K.shape(x)[1]
# if not output_dim:
# # won't work with TensorFlow
# output_dim = K.shape(w)[1]
# if dropout:
# # apply the same dropout pattern at every timestep
# ones = K.ones_like(K.reshape(x[:, 0, :], (-1, input_dim)))
# dropout_matrix = K.dropout(ones, dropout)
# expanded_dropout_matrix = K.repeat(dropout_matrix, timesteps)
# x *= expanded_dropout_matrix
# # collapse time dimension and batch dimension together
# x = K.reshape(x, (-1, input_dim))
# x = K.dot(x, w)
# if b:
# x = x + b
# # reshape to 3D tensor
# x = K.reshape(x, (-1, timesteps, output_dim))
# return x
def load_file(i):
#i = 0
filename = workdir + 'data_combine_code_1D/data_'+str(i+1);
data = pandas.read_csv(filename, engine='python',header=None, sep=r"\s+")
data = data.values
A = []
for i in range(dim_input):
A.append(padding_zero(data[:, i])/normalization_coeff)
Phone = padding_zero_one(data[:,dim_input])
len_data = max_len
Phone = to_categorical(Phone,maxVal_Phone)
A2 = []
for a in A:
A2.append(numpy.reshape(a,(1,len_data,1)))
Phone = numpy.reshape(Phone, (1,len_data, maxVal_Phone))
return A2, Phone
A2, Phone = load_file(0)
for i in range(nb_file-1):
Ak, Phone_k = load_file(i+1)
for i in range(dim_input):
A2[i] = numpy.concatenate((A2[i],Ak[i]), axis=0)
Phone = numpy.concatenate((Phone,Phone_k), axis=0)
def find_len(n):
files_len = []
for i in range(n):
filename = workdir + 'data_combine_code_1D/data_'+str(i+1);
data_in = pandas.read_csv(filename, usecols=[0], engine='python',header=None, sep=r"\s+")
files_len.append(len(data_in))
return files_len
files_len = find_len(nb_file)
def classification_evaluation(Predict,Target):
classification_score = 0
for i in range(len(Predict)):
if int(Predict[i]) == int(Target[i]):
classification_score = classification_score+1
classification_score = classification_score/float(len(Predict))
return classification_score
a_trainScoreClassify1=[]
a_testScoreClassify1=[]
# here is to obtain the random number of the training set
mat = spio.loadmat('/media/liliu/Sony_LiLIU/linux_liliu_beifen/Lip_recognition/random_tab_88.mat', squeeze_me=True)
arr_index = mat['arr_index']
#arr_index = arr_index-1
print(arr_index)
train_A = []
for i in range(dim_input):
train_A.append(A2[i][arr_index[0:nb_file_train]])
train_Phone = Phone[arr_index[0:nb_file_train]]
files_len = numpy.array(files_len)
files_len = files_len[arr_index]
trainX = numpy.concatenate(train_A, axis=2)
input_shape = dim_input
########## build the LSTM model
inputs = Input(shape=(None,input_shape),name = 'inputs')
lstm_out = LSTM(nb_Cell,return_sequences=True,)(inputs)
drop_out = Dropout(drop)(lstm_out)
output1 = Dense(maxVal_Phone,activation="softmax",name='output1')(drop_out)
#output1 = TimeDistributedDense(maxVal_Phone,activation="softmax",name='output1')(drop_out)
#output2 = TimeDistributedDense(maxVal_GT,activation="softmax",name='output2')(drop_out)
#output3 = TimeDistributedDense(maxVal_FX,activation="softmax",name='output3')(drop_out)
#model = Model(input=inputs, output = [output1,output2,output3])
model = Model(input = inputs, output = output1)
#model.compile(optimizer='rmsprop', loss = {'output1':'categorical_crossentropy','output2':'categorical_crossentropy','output3':'categorical_crossentropy'},loss_weights={'output1': 1., 'output2': 1., 'output3': 1.})
model.compile(optimizer = 'rmsprop', loss = 'categorical_crossentropy')
model.fit(trainX,train_Phone, nb_epoch=nb_epoch,batch_size=1, verbose=2)
savefile = workdir + '/model_save/classify_LSTM_1out_nbC_'+str(nb_Cell)+'_nbE_'+str(nb_epoch)+'_nbTf_'+str(nb_file_train)+'.h5';
model.save(savefile)
#model = load_model(savefile)
trainPredict1= model.predict(trainX)
index_trainPredict1 = numpy.argmax(trainPredict1, axis=2)
index_trainTarget1 = numpy.argmax(train_Phone, axis=2)
trainScoreClassify1 = 0
for i in range(len(index_trainPredict1)):
trainScoreClassify1 = trainScoreClassify1+classification_evaluation(index_trainPredict1[i,:], index_trainTarget1[i,:])
trainScoreClassify1 = trainScoreClassify1/float(len(index_trainPredict1));
#for index_filetest in arr_index[nb_file_train:]:
#for index_filetest in range(nb_file_test)+numpy.ones(nb_file_test)*nb_file_train:
for index_filetest in range(nb_file):
test_A = []
for i in range(dim_input):
#len_file_i = files_len[arr_index[index_filetest]]
len_file_i = files_len[index_filetest]
#print(len_file_i)
test_A.append(A2[i][arr_index[index_filetest],0:len_file_i])
test_Phone = Phone[arr_index[index_filetest],0:len_file_i]
test_A[i] = numpy.reshape(test_A[i], (1,len_file_i, 1))
test_Phone = numpy.reshape(test_Phone, (1,len_file_i, maxVal_Phone))
testX = numpy.concatenate(test_A,axis=2)
testPredict1 = model.predict(testX)
index_testPredict1 = numpy.argmax(testPredict1, axis=2)
index_testTarget1 = numpy.argmax(test_Phone, axis=2)
testScoreClassify1 = 0
for i in range(len(index_testPredict1)):
testScoreClassify1 = testScoreClassify1+classification_evaluation(index_testPredict1[i,:], index_testTarget1[i,:])
testScoreClassify1 = testScoreClassify1/float(len(index_testPredict1));
a_testScoreClassify1.append(testScoreClassify1)
a_testScoreClassify1 = numpy.array(a_testScoreClassify1)
print(a_testScoreClassify1)
print(len(a_testScoreClassify1))
avg_testScoreClassify1 = numpy.average(a_testScoreClassify1[nb_file_train:])
print('Train Score: %.2f percent' % (trainScoreClassify1*100))
print('Test Score: %.2f percent' % (avg_testScoreClassify1*100))
with open("classify_LSTM_lip.log", "a") as f:
f.write("nbCell: %d --nbEpoch: %d --nbTrainf: %d --TrScCl1:%f --TeScCl1:%f --Dropout:%f\n" %(nb_Cell,nb_epoch,nb_file_train, trainScoreClassify1, avg_testScoreClassify1,drop))
plt.figure()
plt.plot(a_testScoreClassify1,'ro-')
plt.grid()