forked from xuanjihe/speech-emotion-recognition
-
Notifications
You must be signed in to change notification settings - Fork 0
/
crnn.py
272 lines (230 loc) · 13.5 KB
/
crnn.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
#!/usr/bin/env python2
# -*- coding: utf-8 -*-
"""
Created on Wed Jan 31 16:56:02 2018
@author: hxj
"""
import tensorflow as tf
from tensorflow.python.training import moving_averages
# Importer and Exporting
# ========
tf.app.flags.DEFINE_string ('data_path', './IEMOCAP1.pkl', 'total dataset includes training set, valid set and test set')
tf.app.flags.DEFINE_string ('checkpoint', './checkpoint/', 'the checkpoint dir')
tf.app.flags.DEFINE_string ('model_name', 'model.ckpt', 'model name')
tf.app.flags.DEFINE_string ('pred_name', './pred0.pkl', 'the test output dir')
tf.app.flags.DEFINE_integer ('checkpoint_secs', 60, 'checkpoint saving interval in seconds')
# Global Constants
# ================
tf.app.flags.DEFINE_float ('dropout_conv', 1, 'dropout rate for covvolutional layers')
tf.app.flags.DEFINE_float ('dropout_linear', 1, 'dropout rate for linear layer')
tf.app.flags.DEFINE_float ('dropout_lstm', 1, 'dropout rate for lstm')
tf.app.flags.DEFINE_float ('dropout_fully1', 1, 'dropout rate for fully connected layer1')
tf.app.flags.DEFINE_float ('dropout_fully2', 1, 'dropout rate for fully connected layer1')
#decayed_learning rate
tf.app.flags.DEFINE_float('decay_rate', 0.99, 'the lr decay rate')
tf.app.flags.DEFINE_float('beta1', 0.9, 'parameter of adam optimizer beta1')
tf.app.flags.DEFINE_float('beta2', 0.999, 'adam parameter beta2')
#Moving Average
tf.app.flags.DEFINE_integer('decay_steps', 570, 'the lr decay_step for optimizer')
tf.app.flags.DEFINE_float('momentum', 0.99, 'the momentum')
tf.app.flags.DEFINE_integer('num_epochs', 30000, 'maximum epochs')
tf.app.flags.DEFINE_float ('relu_clip', 20.0, 'ReLU clipping value for non-recurrant layers')
# Adam optimizer (http://arxiv.org/abs/1412.6980) parameters
tf.app.flags.DEFINE_float ('adam_beta1', 0.9, 'beta 1 parameter of Adam optimizer')
tf.app.flags.DEFINE_float ('adam_beta2', 0.999, 'beta 2 parameter of Adam optimizer')
tf.app.flags.DEFINE_float ('epsilon', 1e-8, 'epsilon parameter of Adam optimizer')
tf.app.flags.DEFINE_float ('learning_rate', 0.0001, 'learning rate of Adam optimizer')
# Batch sizes
tf.app.flags.DEFINE_integer ('train_batch_size', 40, 'number of elements in a training batch')
tf.app.flags.DEFINE_integer ('valid_batch_size', 40, 'number of elements in a validation batch')
tf.app.flags.DEFINE_integer ('test_batch_size', 40, 'number of elements in a test batch')
tf.app.flags.DEFINE_integer('save_steps', 10, 'the step to save checkpoint')
tf.app.flags.DEFINE_integer('image_height', 300, 'image height')
tf.app.flags.DEFINE_integer('image_width', 40, 'image width')
tf.app.flags.DEFINE_integer('image_channel', 3, 'image channels as input')
tf.app.flags.DEFINE_integer('linear_num', 786, 'hidden number of linear layer')
tf.app.flags.DEFINE_integer('seq_len', 150, 'sequence length of lstm')
tf.app.flags.DEFINE_integer('cell_num', 128, 'cell units of the lstm')
tf.app.flags.DEFINE_integer('hidden1', 64, 'number of hidden units of fully connected layer')
tf.app.flags.DEFINE_integer('hidden2', 4, 'number of softmax layer')
tf.app.flags.DEFINE_integer('attention_size', 1, 'attention_size')
tf.app.flags.DEFINE_boolean('attention', False, 'whether to use attention, False mean use max-pooling')
FLAGS = tf.app.flags.FLAGS
class CRNN(object):
def __init__(self, mode):
self.mode = mode
# log Mel-spectrogram
self.attention = FLAGS.attention
self.inputs = tf.placeholder(tf.float32, [None, FLAGS.image_height, FLAGS.image_width, FLAGS.image_channel])
# emotion label
self.labels = tf.placeholder(tf.int32, shape=[None, 4])
# lstm time step
#self.seq_len = tf.placeholder(tf.int32, [None])
# l2
self._extra_train_ops = []
def _conv2d(self, x, name, filter_size, in_channels, out_channels, strides):
with tf.variable_scope(name):
kernel = tf.get_variable(name='DW',
shape=[filter_size[0], filter_size[1], in_channels, out_channels],
dtype=tf.float32,
initializer=tf.contrib.layers.xavier_initializer())
b = tf.get_variable(name='bais',
shape=[out_channels],
dtype=tf.float32,
initializer=tf.constant_initializer())
con2d_op = tf.nn.conv2d(x, kernel, [1, strides[0], strides[1], 1], padding='SAME')
return tf.nn.bias_add(con2d_op, b)
def _max_pool(self, x, ksize, strides):
return tf.nn.max_pool(x,
ksize=[1, ksize[0], ksize[1], 1],
strides=[1, strides[0], strides[1], 1],
padding='VALID',
name='max_pool')
def _linear(self,x,names,shapes):
with tf.variable_scope(names):
weights = tf.get_variable(name='weights',
shape=shapes,
initializer=tf.truncated_normal_initializer(stddev=0.1))
bias = tf.get_variable(name='bias',
shape=shapes[1],
initializer=tf.constant_initializer(0.0))
return tf.matmul(x,weights) + bias
def _leaky_relu(self, x, leakiness=0.0):
return tf.where(tf.less(x, 0.0), leakiness * x, x, name='leaky_relu')
def _batch_norm(self, name, x):
"""Batch normalization."""
with tf.variable_scope(name):
params_shape = [x.get_shape()[-1]]
beta = tf.get_variable(
'beta', params_shape, tf.float32,
initializer=tf.constant_initializer(0.0, tf.float32))
gamma = tf.get_variable(
'gamma', params_shape, tf.float32,
initializer=tf.constant_initializer(1.0, tf.float32))
if self.mode == 'train':
mean, variance = tf.nn.moments(x, [0, 1, 2], name='moments')
moving_mean = tf.get_variable(
'moving_mean', params_shape, tf.float32,
initializer=tf.constant_initializer(0.0, tf.float32),
trainable=False)
moving_variance = tf.get_variable(
'moving_variance', params_shape, tf.float32,
initializer=tf.constant_initializer(1.0, tf.float32),
trainable=False)
self._extra_train_ops.append(moving_averages.assign_moving_average(
moving_mean, mean, 0.9))
self._extra_train_ops.append(moving_averages.assign_moving_average(
moving_variance, variance, 0.9))
else:
mean = tf.get_variable(
'moving_mean', params_shape, tf.float32,
initializer=tf.constant_initializer(0.0, tf.float32),
trainable=False)
variance = tf.get_variable(
'moving_variance', params_shape, tf.float32,
initializer=tf.constant_initializer(1.0, tf.float32),
trainable=False)
# tf.summary.histogram(mean.op.name, mean)
# tf.summary.histogram(variance.op.name, variance)
# elipson used to be 1e-5. Maybe 0.001 solves NaN problem in deeper net.
x_bn = tf.nn.batch_normalization(x, mean, variance, beta, gamma, 0.001)
x_bn.set_shape(x.get_shape())
return x_bn
def _batch_norm_wrapper(self, name, inputs, decay = 0.999):
#batch normalization for fully connected layer
with tf.variable_scope(name):
scale = tf.Variable(tf.ones([inputs.get_shape()[-1]]))
beta = tf.Variable(tf.zeros([inputs.get_shape()[-1]]))
pop_mean = tf.Variable(tf.zeros([inputs.get_shape()[-1]]), trainable=False)
pop_var = tf.Variable(tf.ones([inputs.get_shape()[-1]]), trainable=False)
if self.mode == 'train':
batch_mean, batch_var = tf.nn.moments(inputs,[0])
train_mean = tf.assign(pop_mean,
pop_mean * decay + batch_mean * (1 - decay))
train_var = tf.assign(pop_var,
pop_var * decay + batch_var * (1 - decay))
with tf.control_dependencies([train_mean, train_var]):
return tf.nn.batch_normalization(inputs,
batch_mean, batch_var, beta, scale, FLAGS.epsilon)
else:
return tf.nn.batch_normalization(inputs,
pop_mean, pop_var, beta, scale, FLAGS.epsilon)
def _attention(self,inputs, attention_size, time_major=False, return_alphas=False):
if isinstance(inputs, tuple):
# In case of Bi-RNN, concatenate the forward and the backward RNN outputs.
inputs = tf.concat(inputs, 2)
if time_major:
# (T,B,D) => (B,T,D)
inputs = tf.array_ops.transpose(inputs, [1, 0, 2])
hidden_size = inputs.shape[2].value # D value - hidden size of the RNN layer
# Trainable parameters
W_omega = tf.Variable(tf.random_normal([hidden_size, attention_size], stddev=0.1))
b_omega = tf.Variable(tf.random_normal([attention_size], stddev=0.1))
u_omega = tf.Variable(tf.random_normal([attention_size], stddev=0.1))
# Applying fully connected layer with non-linear activation to each of the B*T timestamps;
# the shape of `v` is (B,T,D)*(D,A)=(B,T,A), where A=attention_size
#v = tf.tanh(tf.tensordot(inputs, W_omega, axes=1) + b_omega)
v = tf.sigmoid(tf.tensordot(inputs, W_omega, axes=1) + b_omega)
# For each of the timestamps its vector of size A from `v` is reduced with `u` vector
vu = tf.tensordot(v, u_omega, axes=1) # (B,T) shape
alphas = tf.nn.softmax(vu) # (B,T) shape also
# Output of (Bi-)RNN is reduced with attention vector; the result has (B,D) shape
output = tf.reduce_sum(inputs * tf.expand_dims(alphas, -1), 1)
if not return_alphas:
return output
else:
return output, alphas
def _build_model(self):
filters = [128, 512]
filter_size = [5, 3]
filter_strides = [1, 1]
pool1_size = [2, 4]
pool2_size = [1, 2]
p = 5
with tf.variable_scope('cnn'):
with tf.variable_scope('unit-1'):
x = self._conv2d(self.inputs, 'cnn-1', filter_size, FLAGS.image_channel, filters[0], filter_strides)
x = self._batch_norm('bn1', x)
x = self._leaky_relu(x, 0.01)
x = self._max_pool(x, pool1_size, pool1_size)
# print x.get_shape()
with tf.variable_scope('unit-2'):
x = self._conv2d(x, 'cnn-2', filter_size, filters[0], filters[1], filter_strides)
x = self._batch_norm('bn2', x)
x = self._leaky_relu(x, 0.01)
x = self._max_pool(x, pool2_size, pool2_size)
# print x.get_shape()
with tf.variable_scope('linear'):
# linear layer for dim reduction
x = tf.reshape(x,[-1,p*filters[1]])
x = self._linear(x,'linear1',[p*filters[1],FLAGS.linear_num])
# print x.get_shape()
with tf.variable_scope('lstm'):
x = tf.reshape(x,[-1,FLAGS.seq_len,FLAGS.linear_num])
cell_fw = tf.contrib.rnn.BasicLSTMCell(FLAGS.cell_num, forget_bias=1.0)
if self.mode == 'train':
cell_fw = tf.contrib.rnn.DropoutWrapper(cell=cell_fw, output_keep_prob=FLAGS.dropout_lstm)
cell_bw = tf.contrib.rnn.BasicLSTMCell(FLAGS.cell_num, forget_bias=1.0)
if self.mode == 'train':
cell_bw = tf.contrib.rnn.DropoutWrapper(cell=cell_bw, output_keep_prob=FLAGS.dropout_lstm)
# Now we feed `linear` into the LSTM BRNN cell and obtain the LSTM BRNN output.
outputs, output_states = tf.nn.bidirectional_dynamic_rnn(cell_fw=cell_fw,
cell_bw=cell_bw,
inputs= x,
dtype=tf.float32,
time_major=False,
scope='LSTM1')
with tf.variable_scope('time_pooling'):
if self.attention is not None:
outputs, alphas = self._attention(outputs, FLAGS.attention_size, return_alphas=True)
else:
outputs = tf.concat(outputs,2)
outputs = tf.reshape(outputs, [-1, FLAGS.seq_len,2*FLAGS.cell_num, 1])
outputs = self._max_pool(outputs,[FLAGS.seq_len,1],[FLAGS.seq_len,1])
outputs = tf.reshape(outputs, [-1,2*FLAGS.cell_num])
# print outputs.get_shape()
with tf.variable_scope('dense'):
y = self._linear(outputs,'dense-matmul',[2*FLAGS.cell_num,FLAGS.hidden1])
y = self._batch_norm_wrapper('dense-bn', y)
y = self._leaky_relu(y, 0.01)
self.logits = self._linear(y,'softmax',[FLAGS.hidden1,FLAGS.hidden2])