def predict_save(self):
relabel = relabel()
x_test, test_fname = data_util.load_test()
label_index = [
'down', 'go', 'left', 'no', 'off', 'on', 'right', 'silence',
'stop', 'unknown', 'up', 'yes'
]
print("start predicting...")
predList = []
model = self.fitted_model
if (model == None):
print('model not fitted yet!')
return
predicts = model.predict(x_test)
predicts = np.argmax(predicts, axis=1)
predicts = [label_index[p] for p in predicts]
if not relabel:
predicts = label_transform(predicts,
relabel=True,
get_dummies=False)
df = pd.DataFrame(columns=['fname', 'label'])
df['fname'] = test_fname
df['label'] = predicts
df.to_csv(Configure.submission_path, index=False)
print('done!')
if len(samples) > 16000:
n_samples = chop_audio(samples, num=chopNum)
else:
n_samples = [samples]
for samples in n_samples:
if new_sample_rate >= sample_rate:
resampled = samples
else:
resampled = signal.resample(
samples, int(new_sample_rate / sample_rate * samples.shape[0]))
_, _, specgram = log_specgram(resampled, sample_rate=new_sample_rate)
y_train.append(label)
x_train.append(specgram)
x_train = np.array(x_train)
x_train = x_train.reshape(tuple(list(x_train.shape) + [1]))
y_train = label_transform(y_train, relabel=relabel, get_dummies=True)
label_index = y_train.columns.values
y_train = y_train.values
y_train = np.array(y_train)
del labels, fnames
gc.collect()
print('x_train:', x_train.shape, ', y_train:', y_train.shape)
print("Save train data...")
data_util.save_dataset(x_train, y_train)
del x_train, y_train
gc.collect()
batch = 16
weight = []
#weight = [i/sum(weight) for i in weight]
label_index = [
'down', 'go', 'left', 'no', 'off', 'on', 'right', 'silence', 'stop',
'unknown', 'up', 'yes'
]
modelPaths = glob(os.path.join(Configure.model_path, '*74*'))
modelList = []
for modelFile in modelPaths:
model = load_model(modelFile)
modelList.append(model)
print("start averaging...")
predList = []
for model in modelList:
predicts = model.predict(x_test)
predList.append(predicts)
if len(predList) == len(weight) and sum(weight) == 1.0:
predList = [a * b for a, b in zip(weight, predList)]
predicts = sum(predList) / len(modelPaths)
predicts = np.argmax(predicts, axis=1)
predicts = [label_index[p] for p in predicts]
if not relabel:
predicts = label_transform(predicts, relabel=True, get_dummies=False)
del x_test
gc.collect()
df = pd.DataFrame(columns=['fname', 'label'])
df['fname'] = test_fname
df['label'] = predicts
df.to_csv(Configure.submission_path, index=False)