using a Bi-LSTM layer as the first layer

speech_model.py

@@ -3,7 +3,7 @@ from __future__ import print_function
 import numpy as np
 from speech_data import read_siamese_tfrecords_generator
 from keras.models import Model,load_model,model_from_yaml
-from keras.layers import Input, Dense, Dropout, LSTM, Lambda, Concatenate
+from keras.layers import Input, Dense, Dropout, LSTM, Lambda, Concatenate, Bidirectional
 from keras.losses import categorical_crossentropy
 from keras.utils import to_categorical
 from keras.optimizers import RMSprop
@@ -17,7 +17,7 @@ def create_base_rnn_network(input_dim):
     '''
     inp = Input(shape=input_dim)
     # ls0 = LSTM(512, return_sequences=True)(inp)
-    ls1 = LSTM(256, return_sequences=True)(inp)
+    ls1 = Bidirectional(LSTM(256, return_sequences=True))(inp)
     ls2 = LSTM(128, return_sequences=True)(ls1)
     # ls3 = LSTM(32, return_sequences=True)(ls2)
     ls4 = LSTM(64)(ls2)
@@ -55,10 +55,6 @@ def siamese_model(input_dim):
     processed_b = base_network(input_b)
     final_output = dense_classifier([processed_a,processed_b])
     model = Model([input_a, input_b], final_output)
-    # distance = Lambda(
-    #     euclidean_distance,
-    #     output_shape=eucl_dist_output_shape)([processed_a, processed_b])
-    # model = Model([input_a, input_b], distance)
     return model
 
 def write_model_arch(mod,mod_file):
@@ -73,7 +69,7 @@ def load_model_arch(mod_file):
     return mod
 
 def train_siamese(audio_group = 'audio'):
-    batch_size = 256
+    batch_size = 128
     model_dir = './models/'+audio_group
     create_dir(model_dir)
     log_dir = './logs/'+audio_group