def test_ffnn():
params = {
'n_layers': 4,
'hidden_nodes': [512, 512, 512, 512],
'epochs': 10,
'use_dynamic_features': True,
'use_mspec': False,
'as_mat': False,
'speaker_norm': False,
'context_length': 17
}
net = FFNN(params)
model = net.train_model()
net.set_model(model)
y_true, yp = net.predict_on_test()
print("FFNN RESULTS")
print(get_f1_score(y_true, yp))
print(get_accuracy(y_true, yp))
print(classification_report(y_true, yp))
def test_rnn():
"""Notice as_mat is true here!"""
params = {
'n_layers': 2,
'hidden_nodes': [32, 32],
'epochs': 100,
'use_dynamic_features': True,
'use_mspec': True,
'as_mat': True,
'speaker_norm': False,
'context_length': 35
}
net = RNN(params)
model = net.train_model(params['unroll'])
net.set_model(model)
y_true, yp = net.predict_on_test()
print("RNN RESULTS")
print(get_f1_score(y_true, yp))
print(get_accuracy(y_true, yp))
print(classification_report(y_true, yp))
model.save('rnn-64-64-context-35.h5')
def store_results(y_true, yp, net, model_path, model):
phones = fixed_phones(net)
with open(os.path.join(os.getcwd(), model_path + os.sep + 'results.txt'),
'w') as f:
f.write('acc: {}\n'.format(str(get_accuracy(y_true, yp))))
f.write('edit: {}\n'.format(
str(
eval_edit_dist(y_true,
yp,
net.test,
feature_name=net.feature_name))))
f.write('f1-score: {}\n'.format(str(get_f1_score(y_true, yp))))
report = get_classification_report(y_true, yp, phones)
f.write(str(report))
cm = get_confusion_matrix(y_true, yp)
net.plot_confusion_matrix(
cm,
phones,
os.path.join(os.getcwd(),
model_path + os.sep + 'confusion_matrix.png'),
normalize=True)
model.save(os.path.join(os.getcwd(), model_path + os.sep + 'model.h5'))
def get_f1_score(self):
return metrics.get_f1_score(self.test_labels, self.pred_labels)
def calculate_f1_score(self, samples, model_number):
return get_f1_score(samples, model_number)