malar
/
speech-scoring
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

updated test code

master
Malar Kannan 2017-10-31 17:41:02 +05:30
parent 6fbf06814c
commit 4188585488
2 changed files with 20 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

28
speech_siamese.py
View File

@ -4,8 +4,9 @@ import numpy as np
from speech_data import speech_model_data from speech_data import speech_model_data
from keras.models import Model,load_model from keras.models import Model,load_model
from keras.layers import Input, Dense, Dropout, LSTM, Lambda, Concatenate from keras.layers import Input, Dense, Dropout, LSTM, Lambda, Concatenate
# from keras.losses import categorical_crossentropy from keras.losses import categorical_crossentropy
from keras.losses import binary_crossentropy # from keras.losses import binary_crossentropy
from keras.utils import to_categorical
# from keras.utils.np_utils import to_categorical # from keras.utils.np_utils import to_categorical
from keras.optimizers import RMSprop from keras.optimizers import RMSprop
from keras.callbacks import TensorBoard, ModelCheckpoint from keras.callbacks import TensorBoard, ModelCheckpoint
@ -30,15 +31,14 @@ def contrastive_loss(y_true, y_pred):
return K.mean(y_true * K.square(y_pred) + return K.mean(y_true * K.square(y_pred) +
(1 - y_true) * K.square(K.maximum(1 - y_pred, 0))) (1 - y_true) * K.square(K.maximum(1 - y_pred, 0)))
def create_base_rnn_network(input_dim): def create_base_rnn_network(input_dim):
'''Base network to be shared (eq. to feature extraction). '''Base network to be shared (eq. to feature extraction).
''' '''
inp = Input(shape=input_dim) inp = Input(shape=input_dim)
ls1 = LSTM(1024, return_sequences=True)(inp) ls1 = LSTM(256, return_sequences=True)(inp)
ls2 = LSTM(512, return_sequences=True)(ls1) ls2 = LSTM(128, return_sequences=True)(ls1)
# ls3 = LSTM(32, return_sequences=True)(ls2) # ls3 = LSTM(32, return_sequences=True)(ls2)
ls4 = LSTM(32)(ls2) ls4 = LSTM(64)(ls2)
return Model(inp, ls4) return Model(inp, ls4)
@ -52,15 +52,15 @@ def compute_accuracy(y_true, y_pred):
def accuracy(y_true, y_pred): def accuracy(y_true, y_pred):
'''Compute classification accuracy with a fixed threshold on distances. '''Compute classification accuracy with a fixed threshold on distances.
''' '''
return K.mean(K.equal(y_true, K.cast(y_pred < 0.5, y_true.dtype))) return K.mean(K.equal(y_true, K.cast(y_pred > 0.5, y_true.dtype)))
def dense_classifier(processed): def dense_classifier(processed):
conc_proc = Concatenate()(processed) conc_proc = Concatenate()(processed)
d1 = Dense(8, activation='relu')(conc_proc) d1 = Dense(16, activation='relu')(conc_proc)
dr1 = Dropout(0.1)(d1) # dr1 = Dropout(0.1)(d1)
# d2 = Dense(8, activation='relu')(dr1) d2 = Dense(8, activation='relu')(d1)
# dr2 = Dropout(0.1)(d2) # dr2 = Dropout(0.1)(d2)
return Dense(1, activation='sigmoid')(dr1) return Dense(2, activation='softmax')(d2)
def siamese_model(input_dim): def siamese_model(input_dim):
# input_dim = (15, 1654) # input_dim = (15, 1654)
@ -80,7 +80,9 @@ def siamese_model(input_dim):
def train_siamese(): def train_siamese():
# the data, shuffled and split between train and test sets # the data, shuffled and split between train and test sets
tr_pairs, te_pairs, tr_y, te_y = speech_model_data() tr_pairs, te_pairs, tr_y_e, te_y_e = speech_model_data()
tr_y = to_categorical(tr_y_e, num_classes=2)
te_y = to_categorical(te_y_e, num_classes=2)
input_dim = (tr_pairs.shape[2], tr_pairs.shape[3]) input_dim = (tr_pairs.shape[2], tr_pairs.shape[3])
model = siamese_model(input_dim) model = siamese_model(input_dim)
@ -108,7 +110,7 @@ def train_siamese():
period=1) period=1)
# train # train
rms = RMSprop(lr=0.001) rms = RMSprop(lr=0.001)
model.compile(loss=binary_crossentropy, optimizer=rms, metrics=[accuracy]) model.compile(loss=categorical_crossentropy, optimizer=rms, metrics=[accuracy])
model.fit( model.fit(
[tr_pairs[:, 0], tr_pairs[:, 1]], [tr_pairs[:, 0], tr_pairs[:, 1]],
tr_y, tr_y,

7
test_siamese.py
View File

@ -5,6 +5,7 @@ import speech_data
reload(speech_data) reload(speech_data)
from speech_data import create_test_pair,get_word_pairs_data from speech_data import create_test_pair,get_word_pairs_data
import numpy as np import numpy as np
from keras.utils import to_categorical
model = siamese_model((15, 1654)) model = siamese_model((15, 1654))
model.load_weights('./models/siamese_speech_model-final.h5') model.load_weights('./models/siamese_speech_model-final.h5')
@ -17,6 +18,6 @@ def predict_recording_with(m,sample_size=15):
predict_recording_with(model) predict_recording_with(model)
sunflower_data,sunflower_result = get_word_pairs_data('sunflowers',15) sunflower_data,sunflower_result = get_word_pairs_data('sweater',15)
sunflower_result print(np.argmax(model.predict([sunflower_data[:, 0], sunflower_data[:, 1]]),axis=1))
model.predict([sunflower_data[:, 0], sunflower_data[:, 1]]) < 0.5 print(sunflower_result)