def test(file_name):
demo_file = file_name
# demo=data.load_wav_file(data.path + demo_file)
demo = data.load_wav_file("data/test/" + demo_file)
result = model.predict([demo])
conf = numpy.amax(result) * 100
result = data.one_hot_to_item(result, speakers)
print("predicted speaker for %s : result = %s confidence = %.2f" %
(demo_file, result, conf))
def validateWav(demo_file):
demoData = speech_data.load_wav_file(
speech_data.snore_train_path + demo_file, 0)
result = model.predict([demoData])
if (result[0][1] > 0.6):
rc = 1
print(demo_file + ":" +
str((result[0][0])) + " / " + str((result[0][1])) + ": ISSNORE")
else:
print(demo_file + ":" +
str((result[0][0])) + " / " + str((result[0][1])))
rc = 0
return rc
def handle_speaker_rec_test_intent(self, message):
speakers = data.get_speakers()
number_classes = len(speakers)
#print("speakers",speakers)
#batch=data.wave_batch_generator(batch_size=1000, source=data.Source.DIGIT_WAVES, target=data.Target.speaker)
#X,Y=next(batch)
# Classification
#tflearn.init_graph(num_cores=8, gpu_memory_fraction=0.5)
net = tflearn.input_data(shape=[None, 8192]) #Two wave chunks
net = tflearn.fully_connected(net, 64)
net = tflearn.dropout(net, 0.5)
net = tflearn.fully_connected(net,
number_classes,
activation='softmax')
net = tflearn.regression(net,
optimizer='adam',
loss='categorical_crossentropy')
model = tflearn.DNN(net)
#model.fit(X, Y, n_epoch=100, show_metric=True, snapshot_step=100)
CWD_PATH = os.path.dirname(__file__)
path_to_model = os.path.join(CWD_PATH, 'model', 'model.tfl')
model.load(path_to_model)
demo_file = "8_Vicki_260.wav"
#demo_file = "8_Bruce_260.wav"
demo = data.load_wav_file(data.path + demo_file)
result = model.predict([demo])
result = data.one_hot_to_item(result, speakers)
if result == "Vicki":
self.speak("I am confident I'm speaking to %s" %
(result)) # ~ 97% correct
else:
self.speak("I'm sorry I don't recognize your voice")
number_classes = 10 # Digits
# Classification
tflearn.init_graph(num_cores=8, gpu_memory_fraction=0.5)
net = tflearn.input_data(shape=[None, 8192])
#net = tflearn.fully_connected(net, 64)
net = tflearn.fully_connected(net, 64)
net = tflearn.dropout(net, 0.5)
net = tflearn.fully_connected(net, number_classes, activation='softmax')
net = tflearn.regression(net,
optimizer='adam',
loss='categorical_crossentropy')
model = tflearn.DNN(net)
model.fit(X, Y, n_epoch=5, show_metric=True, snapshot_step=100)
# Overfitting okay for now
print("================= TEST 결과 ================")
path = "data/number_test/"
files = os.listdir(path)
print("loaded Test batch of %d files" % len(files))
for wav in files:
demo_file = path + wav
demo = speech_data.load_wav_file(demo_file)
result = model.predict([demo])
result = numpy.argmax(result)
print("== predicted digit for %s :result = %d " % (demo_file, result))
number_classes = len(speakers)
##print("speakers",speakers)
WORD_WAVs = "spoken_words"
batch = data.wave_batch_generator(batch_size=1000,
source=WORD_WAVs,
target=data.Target.speaker)
X, Y = next(batch)
# Classification
tflearn.init_graph(num_cores=8, gpu_memory_fraction=0.5)
net = tflearn.input_data(shape=[None, 8192]) #Two wave chunks
net = tflearn.fully_connected(net, 64)
net = tflearn.dropout(net, 0.5)
net = tflearn.fully_connected(net, number_classes, activation='softmax')
net = tflearn.regression(net,
optimizer='adam',
loss='categorical_crossentropy')
model = tflearn.DNN(net)
##model.fit(X, Y, n_epoch=100, show_metric=True, snapshot_step=100)
model.load("tflearn.dnn.model")
# demo_file = "8_Vicki_260.wav"
demo_file = "8_Bruce_260.wav"
demo = data.load_wav_file(data.path + demo_file)
result = model.predict([demo])
result = data.one_hot_to_item(result, speakers)
model.save("tflearn.lstm.model")
print("predicted speaker for %s : result = %s " % (demo_file, result))
# Training Step: 544 | total loss: 0.15866
# | Adam | epoch: 034 | loss: 0.15866 - acc: 0.9818 -- iter: 0000/1000
batch = speech_data.wave_batch_generator(10000,
target=speech_data.Target.digits)
X, Y = next(batch)
number_classes = 10 # Digits
# Classification
tflearn.init_graph(num_cores=8, gpu_memory_fraction=0.5)
net = tflearn.input_data(shape=[None, 8192])
net = tflearn.fully_connected(net, 64)
net = tflearn.dropout(net, 0.5)
net = tflearn.fully_connected(net, number_classes, activation='softmax')
net = tflearn.regression(net,
optimizer='adam',
loss='categorical_crossentropy')
model = tflearn.DNN(net)
model.fit(X, Y, n_epoch=3, show_metric=True, snapshot_step=100)
# Overfitting okay for now
demo_file = "5_Vicki_260.wav"
demo = speech_data.load_wav_file(speech_data.path + demo_file)
result = model.predict([demo])
result = numpy.argmax(result)
print("predicted digit for %s : result = %d " % (demo_file, result))
import numpy
# Simple spoken digit recognition demo, with 98% accuracy in under a minute
# Training Step: 544 | total loss: 0.15866
# | Adam | epoch: 034 | loss: 0.15866 - acc: 0.9818 -- iter: 0000/1000
batch=speech_data.wave_batch_generator(10000,target=speech_data.Target.digits)
X,Y=next(batch)
number_classes=10 # Digits
# Classification
tflearn.init_graph(num_cores=8, gpu_memory_fraction=0.5)
net = tflearn.input_data(shape=[None, 8192])
net = tflearn.fully_connected(net, 64)
net = tflearn.dropout(net, 0.5)
net = tflearn.fully_connected(net, number_classes, activation='softmax')
net = tflearn.regression(net, optimizer='adam', loss='categorical_crossentropy')
model = tflearn.DNN(net)
model.fit(X, Y,n_epoch=3,show_metric=True,snapshot_step=100)
# Overfitting okay for now
demo_file = "5_Vicki_260.wav"
demo=speech_data.load_wav_file(speech_data.path + demo_file)
result=model.predict([demo])
result=numpy.argmax(result)
print("predicted digit for %s : result = %d "%(demo_file,result))
# | Adam | epoch: 030 | loss: 0.05330 - acc: 0.9966 -- iter: 0000/1000
# 'predicted speaker for 9_Vicki_260 : result = ', 'Vicki'
speakers = data.get_speakers()
number_classes=len(speakers)
print("speakers",speakers)
WORD_WAVs="spoken_words"
batch=data.wave_batch_generator(batch_size=1000,source=WORD_WAVs,target=data.Target.speaker)
X,Y=next(batch)
# Classification
tflearn.init_graph(num_cores=8, gpu_memory_fraction=0.5)
net = tflearn.input_data(shape=[None, 8192]) #Two wave chunks
net = tflearn.fully_connected(net, 64)
net = tflearn.dropout(net, 0.5)
net = tflearn.fully_connected(net, number_classes, activation='softmax')
net = tflearn.regression(net, optimizer='adam', loss='categorical_crossentropy')
model = tflearn.DNN(net)
model.fit(X, Y, n_epoch=100, show_metric=True, snapshot_step=100)
# demo_file = "8_Vicki_260.wav"
demo_file = "8_Bruce_260.wav"
demo=data.load_wav_file(data.path + demo_file)
result=model.predict([demo])
result=data.one_hot_to_item(result,speakers)
print("predicted speaker for %s : result = %s "%(demo_file,result))
optimizer='adam',
loss='categorical_crossentropy')
model = tflearn.DNN(regression)
model.fit(X, Y, n_epoch=200, show_metric=True, snapshot_step=100)
model.save("model.tfl")
# model.load("model.tfl")
m2 = tflearn.DNN(fc2, session=model.session)
baza_path = "data/baza_glosow/"
baza_path_vect = "data/baza_glosow/vect/"
# Feature extraction phase
for filename in os.listdir(baza_path):
if filename.endswith(".wav"):
demo = data.load_wav_file(baza_path + filename)
vect = m2.predict([demo])
sq_vect = np.squeeze(vect)
np.savetxt(baza_path_vect + filename + ".npz", sq_vect, delimiter=',')
# Verification phase
cnt = 0
sum = 0
for filename in os.listdir(baza_path):
if filename.endswith(".wav"):
out = cluster_vector.cluster(baza_path_vect,
os.listdir(baza_path_vect), filename[:-4])
print(filename)
print(out)
correct_cnt = 0
for it in out:
# Training Step: 544 | total loss: 0.15866
# | Adam | epoch: 034 | loss: 0.15866 - acc: 0.9818 -- iter: 0000/1000
if __name__ == '__main__':
batch = wave_batch_generator(10000, target=Target.digits)
X, Y = next(batch)
number_classes = 10 # Digits
# Classification
tflearn.init_graph(num_cores=8, gpu_memory_fraction=0.5)
net = tflearn.input_data(shape=[None, 8192])
net = tflearn.fully_connected(net, 64)
net = tflearn.dropout(net, 0.5)
net = tflearn.fully_connected(net, number_classes, activation='softmax')
net = tflearn.regression(net,
optimizer='adam',
loss='categorical_crossentropy')
model = tflearn.DNN(net)
model.fit(X, Y, n_epoch=3, show_metric=True, snapshot_step=100)
# Overfitting okay for now
demo_file = "5_Vicki_260.wav"
demo = load_wav_file(path + demo_file)
result = model.predict([demo])
result = numpy.argmax(result)
print("predicted digit for %s : result = %d " % (demo_file, result))
tflearn.init_graph(num_cores=8, gpu_memory_fraction=0.5)
net = tflearn.input_data(shape=[None, 8192])
net = tflearn.fully_connected(net, 64)
net = tflearn.dropout(net, 0.5)
net = tflearn.fully_connected(net, number_classes, activation='softmax')
net = tflearn.regression(net, optimizer='adam', loss='categorical_crossentropy')
model = tflearn.DNN(net)
model.fit(X, Y,n_epoch=10,show_metric=True,snapshot_step=100)
model.save("model.tfm")
# model.load("model.tfm")
baza_path = "data/spoken_numbers_pcm/tt/"
cnt = 0
sum = 0
for filename in os.listdir(baza_path):
if filename.endswith(".wav"):
demo=speech_data.load_wav_file(baza_path + filename)
result=model.predict([demo])
result=numpy.argmax(result)
print("predicted speaker for %s : result = %d "%(filename, result))
out = cluster_vector.cluster(baza_path_vect, os.listdir(baza_path_vect), filename[:-4])
correct_cnt = 0
for it in out:
if it==''.join(filter(str.islower, filename[:-4])):
correct_cnt += 1
sum += (correct_cnt-1)/2
cnt +=1
print(sum)
