def train(args):
src_root = args.src_root
sr = args.sample_rate
dt = args.delta_time
batch_size = args.batch_size
model_type = args.model_type
params = {'N_CLASSES': len(os.listdir(args.src_root)), 'SR': sr, 'DT': dt}
models = {
'conv1d': Conv1D(**params),
'conv2d': Conv2D(**params),
'lstm': LSTM(**params)
}
assert model_type in models.keys(), '{} not an available model'.format(
model_type)
csv_path = os.path.join('logs', '{}_history.csv'.format(model_type))
wav_paths = glob('{}/**'.format(src_root), recursive=True)
wav_paths = [x.replace(os.sep, '/') for x in wav_paths if '.wav' in x]
classes = sorted(os.listdir(args.src_root))
le = LabelEncoder()
le.fit(classes)
labels = [os.path.split(x)[0].split('/')[-1] for x in wav_paths]
labels = le.transform(labels)
wav_train, wav_val, label_train, label_val = train_test_split(
wav_paths, labels, test_size=0.1, random_state=0)
assert len(
label_train
) >= args.batch_size, 'Anzahl der trainierten Dateien muss >= batch_size sein'
tg = DataGenerator(wav_train,
label_train,
sr,
dt,
len(set(label_train)),
batch_size=batch_size)
vg = DataGenerator(wav_val,
label_val,
sr,
dt,
len(set(label_val)),
batch_size=batch_size)
model = models[model_type]
cp = ModelCheckpoint('models/{}.h5'.format(model_type),
monitor='val_loss',
save_best_only=True,
save_weights_only=False,
mode='auto',
save_freq='epoch',
verbose=1)
csv_logger = CSVLogger(csv_path, append=False)
model.fit(tg,
validation_data=vg,
epochs=30,
verbose=1,
callbacks=[csv_logger, cp])
def train(args):
src_root = args.src_root
sr = args.sample_rate
dt = args.delta_time
batch_size = args.batch_size
model_type = args.model_type
models = {'conv1d': Conv1D(), 'conv2d': Conv2D(), 'lstm': LSTM()}
assert model_type in models.keys(), '{} not an available model'.format(
model_type)
wav_paths = glob('{}/**'.format(src_root), recursive=True)
wav_paths = [x for x in wav_paths if '.wav' in x]
classes = sorted(os.listdir(args.src_root))
le = LabelEncoder()
le.fit(classes)
labels = [get_class(x, src_root) for x in wav_paths]
labels = le.transform(labels)
wav_train, wav_val, label_train, label_val = train_test_split(
wav_paths, labels, test_size=0.1, random_state=0)
tg = DataGenerator(wav_train,
label_train,
sr,
dt,
len(set(label_train)),
batch_size=batch_size)
vg = DataGenerator(wav_val,
label_val,
sr,
dt,
len(set(label_val)),
batch_size=batch_size)
model = models[model_type]
cp = ModelCheckpoint('models/{}.h5'.format(model_type),
monitor='val_loss',
save_best_only=True,
save_weights_only=False,
mode='auto',
save_freq='epoch',
verbose=1)
csv_logger = CSVLogger(os.path.join('logs',
'{}_history.csv'.format(model_type)),
append=True)
model.fit(tg,
validation_data=vg,
epochs=30,
verbose=1,
callbacks=[csv_logger, cp])
def train(args):
src_root = args.src_root
sr = args.sample_rate
dt = args.delta_time
batch_size = args.batch_size
model_type = args.model_type
params = {'N_CLASSES':len(os.listdir(args.src_root)),
'SR':sr,
'DT':dt}
models = {'conv1d':Conv1D(**params),
'conv2d':Conv2D(**params),
'lstm': LSTM(**params),
'transformer': Transformer(**params),
'ViT': ViT(**params)}
assert model_type in models.keys(), '{} not an available model'.format(model_type)
csv_path = os.path.join('logs', '{}_history.csv'.format(model_type))
wav_paths = glob('{}/**'.format(src_root), recursive=True)
wav_paths = [x.replace(os.sep, '/') for x in wav_paths if '.wav' in x]
classes = sorted(os.listdir(args.src_root))
le = LabelEncoder()
le.fit(classes)
labels = [os.path.split(x)[0].split('/')[-1] for x in wav_paths]
labels = le.transform(labels)
wav_train, wav_val, label_train, label_val = train_test_split(wav_paths,
labels,
test_size=0.1,
random_state=0)
assert len(label_train) >= args.batch_size, 'Number of train samples must be >= batch_size'
if len(set(label_train)) != params['N_CLASSES']:
warnings.warn('Found {}/{} classes in training data. Increase data size or change random_state.'.format(len(set(label_train)), params['N_CLASSES']))
if len(set(label_val)) != params['N_CLASSES']:
warnings.warn('Found {}/{} classes in validation data. Increase data size or change random_state.'.format(len(set(label_val)), params['N_CLASSES']))
tg = DataGenerator(wav_train, label_train, sr, dt,
params['N_CLASSES'], batch_size=batch_size)
vg = DataGenerator(wav_val, label_val, sr, dt,
params['N_CLASSES'], batch_size=batch_size)
model = models[model_type]
model.summary()
cp = ModelCheckpoint('models/{}.h5'.format(model_type), monitor='val_loss',
save_best_only=True, save_weights_only=False,
mode='auto', save_freq='epoch', verbose=1)
csv_logger = CSVLogger(csv_path, append=False)
model.fit(tg, validation_data=vg,
epochs=40, verbose=1,
callbacks=[csv_logger])
def main(args):
with open(args.config, 'r') as f:
config = json.load(f)
if config['optimizer'] == 'SGD':
optimizer = SGD(lr=config['learning_rate'],
decay=config['learning_rate'] / config['epochs'],
momentum=config['momentum'])
else:
raise Exception('Unsupported optimizer: {}.'.format(
config['optimizer']))
model_name = str.lower(config['model'])
if model_name == 'lstm':
model = LSTM(config['input_length'], 2)
elif model_name == 'conv1d':
model = Conv1D(config['input_length'], 2)
elif model_name == 'conv2d':
model = Conv2D(config['input_length'], 2)
else:
raise Exception('Unsupported model: {}.'.format(config['model']))
model.compile(loss='binary_crossentropy',
optimizer=optimizer,
metrics=['accuracy'])
wav_paths = glob('{}/**'.format(args.data_dir), recursive=True)
wav_paths = [x for x in wav_paths if '.wav' in x]
classes = sorted(os.listdir(args.data_dir))
le = LabelEncoder()
le.fit(classes)
labels = [get_class(x, args.data_dir) for x in wav_paths]
labels = le.transform(labels)
print('CLASSES: ', list(le.classes_))
print(le.transform(list(le.classes_)))
wav_train, wav_val, label_train, label_val = train_test_split(
wav_paths,
labels,
test_size=config['validation_split'],
random_state=0)
tg = DataGenerator(wav_train,
label_train,
config['input_length'],
len(set(label_train)),
batch_size=config['batch_size'])
vg = DataGenerator(wav_val,
label_val,
config['input_length'],
len(set(label_val)),
batch_size=config['batch_size'])
output_sub_dir = os.path.join(args.output_dir, model_name,
datetime.now().strftime('%Y%m%d_%H%M%S'))
os.makedirs(output_sub_dir)
callbacks = [
EarlyStopping(monitor='val_loss',
patience=config['patience'],
restore_best_weights=True,
verbose=1),
ModelCheckpoint(filepath=os.path.join(
output_sub_dir, 'model.{epoch:02d}-{val_loss:.4f}.h5'),
monitor='val_loss',
save_best_only=True,
verbose=1),
CSVLogger(os.path.join(output_sub_dir, 'epochs.csv'))
]
model.fit(tg,
validation_data=vg,
epochs=config['epochs'],
verbose=1,
callbacks=callbacks)