-
Notifications
You must be signed in to change notification settings - Fork 180
/
arabic_voice_model.py
201 lines (141 loc) · 5.6 KB
/
arabic_voice_model.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
from keras.models import Model
from keras.layers import Input, Dense, LSTM, multiply, concatenate, Activation, Masking, Reshape
from keras.layers import Conv1D, BatchNormalization, GlobalAveragePooling1D, Permute, Dropout
from keras.regularizers import l2
from utils.constants import MAX_NB_VARIABLES, NB_CLASSES_LIST, MAX_TIMESTEPS_LIST
from utils.keras_utils import train_model, evaluate_model, set_trainable
from utils.layer_utils import AttentionLSTM
DATASET_INDEX = 5
MAX_TIMESTEPS = MAX_TIMESTEPS_LIST[DATASET_INDEX]
MAX_NB_VARIABLES = MAX_NB_VARIABLES[DATASET_INDEX]
NB_CLASS = NB_CLASSES_LIST[DATASET_INDEX]
TRAINABLE = True
regularization_weight = 5e-4
def generate_model():
ip = Input(shape=(MAX_NB_VARIABLES, MAX_TIMESTEPS))
x = Masking()(ip)
x = LSTM(8)(x)
x = Dropout(0.8)(x)
y = Permute((2, 1))(ip)
y = Conv1D(128, 8, padding='same', kernel_initializer='he_uniform')(y)
y = BatchNormalization()(y)
y = Activation('relu')(y)
y = squeeze_excite_block(y)
y = Conv1D(256, 5, padding='same', kernel_initializer='he_uniform')(y)
y = BatchNormalization()(y)
y = Activation('relu')(y)
y = squeeze_excite_block(y)
y = Conv1D(128, 3, padding='same', kernel_initializer='he_uniform')(y)
y = BatchNormalization()(y)
y = Activation('relu')(y)
y = GlobalAveragePooling1D()(y)
x = concatenate([x, y])
out = Dense(NB_CLASS, activation='softmax')(x)
model = Model(ip, out)
model.summary()
# add load model code here to fine-tune
return model
def generate_model_2():
ip = Input(shape=(MAX_NB_VARIABLES, MAX_TIMESTEPS))
# stride = 10
# x = Permute((2, 1))(ip)
# x = Conv1D(MAX_NB_VARIABLES // stride, 8, strides=stride, padding='same', activation='relu', use_bias=False,
# kernel_initializer='he_uniform')(x) # (None, variables / stride, timesteps)
# x = Permute((2, 1))(x)
#ip1 = K.reshape(ip,shape=(MAX_TIMESTEPS,MAX_NB_VARIABLES))
#x = Permute((2, 1))(ip)
x = Masking()(ip)
x = AttentionLSTM(8)(x)
x = Dropout(0.8)(x)
y = Permute((2, 1))(ip)
y = Conv1D(128, 8, padding='same', kernel_initializer='he_uniform')(y)
y = BatchNormalization()(y)
y = Activation('relu')(y)
y = squeeze_excite_block(y)
y = Conv1D(256, 5, padding='same', kernel_initializer='he_uniform')(y)
y = BatchNormalization()(y)
y = Activation('relu')(y)
y = squeeze_excite_block(y)
y = Conv1D(128, 3, padding='same', kernel_initializer='he_uniform')(y)
y = BatchNormalization()(y)
y = Activation('relu')(y)
y = GlobalAveragePooling1D()(y)
x = concatenate([x, y])
out = Dense(NB_CLASS, activation='softmax')(x)
model = Model(ip, out)
model.summary()
# add load model code here to fine-tune
return model
def generate_model_3():
ip = Input(shape=(MAX_NB_VARIABLES, MAX_TIMESTEPS))
x = Masking()(ip)
x = LSTM(8)(x)
x = Dropout(0.8)(x)
y = Permute((2, 1))(ip)
y = Conv1D(128, 8, padding='same', kernel_initializer='he_uniform')(y)
y = BatchNormalization()(y)
y = Activation('relu')(y)
#y = squeeze_excite_block(y)
y = Conv1D(256, 5, padding='same', kernel_initializer='he_uniform')(y)
y = BatchNormalization()(y)
y = Activation('relu')(y)
#y = squeeze_excite_block(y)
y = Conv1D(128, 3, padding='same', kernel_initializer='he_uniform')(y)
y = BatchNormalization()(y)
y = Activation('relu')(y)
y = GlobalAveragePooling1D()(y)
x = concatenate([x, y])
out = Dense(NB_CLASS, activation='softmax')(x)
model = Model(ip, out)
model.summary()
# add load model code here to fine-tune
return model
def generate_model_4():
ip = Input(shape=(MAX_NB_VARIABLES, MAX_TIMESTEPS))
# stride = 3
#
# x = Permute((2, 1))(ip)
# x = Conv1D(MAX_NB_VARIABLES // stride, 8, strides=stride, padding='same', activation='relu', use_bias=False,
# kernel_initializer='he_uniform')(x) # (None, variables / stride, timesteps)
# x = Permute((2, 1))(x)
x = Masking()(ip)
x = AttentionLSTM(8)(x)
x = Dropout(0.8)(x)
y = Permute((2, 1))(ip)
y = Conv1D(128, 8, padding='same', kernel_initializer='he_uniform')(y)
y = BatchNormalization()(y)
y = Activation('relu')(y)
#y = squeeze_excite_block(y)
y = Conv1D(256, 5, padding='same', kernel_initializer='he_uniform')(y)
y = BatchNormalization()(y)
y = Activation('relu')(y)
#y = squeeze_excite_block(y)
y = Conv1D(128, 3, padding='same', kernel_initializer='he_uniform')(y)
y = BatchNormalization()(y)
y = Activation('relu')(y)
y = GlobalAveragePooling1D()(y)
x = concatenate([x, y])
out = Dense(NB_CLASS, activation='softmax')(x)
model = Model(ip, out)
model.summary()
# add load model code here to fine-tune
return model
def squeeze_excite_block(input):
''' Create a squeeze-excite block
Args:
input: input tensor
filters: number of output filters
k: width factor
Returns: a keras tensor
'''
filters = input._keras_shape[-1] # channel_axis = -1 for TF
se = GlobalAveragePooling1D()(input)
se = Reshape((1, filters))(se)
se = Dense(filters // 16, activation='relu', kernel_initializer='he_normal', use_bias=False)(se)
se = Dense(filters, activation='sigmoid', kernel_initializer='he_normal', use_bias=False)(se)
se = multiply([input, se])
return se
if __name__ == "__main__":
model = generate_model_2()
# train_model(model, DATASET_INDEX, dataset_prefix='arabic_voice', epochs=600, batch_size=128)
evaluate_model(model, DATASET_INDEX, dataset_prefix='arabic_voice', batch_size=128)