"Dataset not registered. Please create a method to read it")
db = Dataset(path, dataset, decode=False)
print("Dumping " + dataset + " data set to file...")
cPickle.dump(db, open(dataset + '_db.p', 'wb'))
else:
print("Loading data from " + dataset + " data set...")
db = cPickle.load(open(dataset + '_db.p', 'rb'))
nb_samples = len(db.targets)
print("Number of samples: " + str(nb_samples))
if feature_extract:
f_global = functions.feature_extract(db.data,
nb_samples=nb_samples,
dataset=dataset)
else:
print("Loading features from file...")
f_global = cPickle.load(open(dataset + '_features.p', 'rb'))
y = np.array(db.targets)
y = to_categorical(y, num_classes=globalvars.nb_classes)
if speaker_independence:
k_folds = len(db.test_sets)
splits = zip(db.train_sets, db.test_sets)
print("Using speaker independence %s-fold cross validation" % k_folds)
else:
k_folds = 10
sss = StratifiedShuffleSplit(n_splits=k_folds,
globalvars.dataset = dataset
globalvars.nb_classes = nb_classes
if load_data:
ds = Dataset(path=path, dataset=dataset)
print("Writing " + dataset + " data set to file...")
cPickle.dump(ds, open(dataset + '_db.p', 'wb'))
else:
print("Loading data from " + dataset + " data set...")
ds = cPickle.load(open(dataset + '_db.p', 'rb'))
if feature_extract:
functions.feature_extract(ds.data,
nb_samples=len(ds.targets),
dataset=dataset)
try:
trials = Trials()
best_run, best_model = optim.minimize(model=create_model,
data=get_data,
algo=tpe.suggest,
max_evals=6,
trials=trials)
U_train, X_train, Y_train, U_test, X_test, Y_test = get_data()
best_model_idx = 1
best_score = 0.0
for i in range(1, (globalvars.globalVar + 1)):
parser.add_option('-c', '--nb_classes', dest='nb_classes', type='int', default=7)
(options, args) = parser.parse_args(sys.argv)
wav_path = options.wav_path
load_data = options.load_data
feature_extract = options.feature_extract
model_path = options.model_path
nb_classes = options.nb_classes
globalvars.nb_classes = nb_classes
y, sr = librosa.load(wav_path, sr=16000)
wav = AudioSegment.from_file(wav_path)
if feature_extract:
f = functions.feature_extract((y, sr), nb_samples=1, dataset='prediction')
else:
print("Loading features from file...")
f = cPickle.load(open('prediction_features.p', 'rb'))
u = np.full((f.shape[0], globalvars.nb_attention_param), globalvars.attention_init_value,
dtype=np.float64)
# load model
model = load_model(model_path)
# prediction
results = model.predict([u, f], batch_size=128, verbose=1)
for result in results:
print(result)