def __init__(self,
path_model='',
path_dataset='',
ae_type='conv_ae',
name='conv_ae',
optimizer='adam',
shape=(60, 56),
beta=3,
rho=0.005):
self.ae_type = ae_type
self.name = name
self.path_model = path_model
if path_model != '':
self.path_dataset = path_dataset
self.path_autoencoder = self.path_model + '/autoencoder.h5'
self.path_summary = self.path_model + '/summary.txt'
self.path_loss_progress = self.path_model + '/training.log'
self.shape = shape
self.autoencoder_model = load_model(self.path_autoencoder)
else:
self.path_dataset = path_dataset
self.base_dir = model_utils.get_ae_base_dir(self.ae_type, self.name)
self.path_autoencoder = self.base_dir + '/autoencoder.h5'
self.path_summary = self.base_dir + '/summary.txt'
self.path_csv_logger = self.base_dir + '/training.log'
self.shape = shape
self.autoencoder_model = self.build_model()
self.autoencoder_model.compile(loss='mse', optimizer=optimizer)
def __init__(self, path_model='', path_dataset='', name='deep_ae', encoding_dim=8):
self.name = name
self.path_model = path_model
if path_model != '':
self.path_autoencoder = self.path_model + '/autoencoder.h5'
self.path_summary = self.path_model + '/summary.txt'
self.path_loss_progress = self.path_model + '/training.log'
self.autoencoder_model = load_model(self.path_autoencoder)
self.dataset_config = model_utils.get_dataset_config(self.path_model)
self.path_dataset = self.dataset_config['PATH_DATASET']
self.original_shape = self.dataset_config['ORIGINAL_SHAPE']
self.shape = np.prod(self.original_shape)
else:
self.path_dataset = path_dataset
self.base_dir = model_utils.get_ae_base_dir(self.name)
self.path_autoencoder = self.base_dir + '/autoencoder.h5'
self.path_summary = self.base_dir + '/summary.txt'
self.path_csv_logger = self.base_dir + '/training.log'
self.dataset_config = model_utils.get_dataset_config(self.path_dataset)
model_utils.copy_dataset_config(self.path_dataset, self.base_dir)
self.original_shape = self.dataset_config['ORIGINAL_SHAPE']
self.shape = np.prod(self.original_shape)
self.encoding_dim = encoding_dim
optimizer = Adam(lr=0.0001)
self.autoencoder_model = self.build_model()
self.autoencoder_model.compile(loss='mse', optimizer=optimizer)
def __init__(self,
path_model='',
path_dataset='',
name='conv_kl_ae',
beta=3,
rho=0.005):
self.name = name
self.path_model = path_model
if path_model != '':
self.path_autoencoder = self.path_model + '/autoencoder.h5'
self.path_summary = self.path_model + '/summary.txt'
self.path_loss_progress = self.path_model + '/training.log'
custom_objects = {'SparsityRegularizer': SparsityRegularizer}
self.autoencoder_model = load_model(self.path_autoencoder,
custom_objects=custom_objects)
self.dataset_config = model_utils.get_dataset_config(
self.path_model)
self.path_dataset = self.dataset_config['PATH_DATASET']
self.shape = self.dataset_config['ORIGINAL_SHAPE']
else:
self.path_dataset = path_dataset
self.base_dir = model_utils.get_ae_base_dir(self.name)
self.path_autoencoder = self.base_dir + '/autoencoder.h5'
self.path_summary = self.base_dir + '/summary.txt'
self.path_csv_logger = self.base_dir + '/training.log'
self.dataset_config = model_utils.get_dataset_config(
self.path_dataset)
model_utils.copy_dataset_config(self.path_dataset, self.base_dir)
self.shape = self.dataset_config['ORIGINAL_SHAPE']
optimizer = Adam(lr=0.0001)
self.rho = rho
self.beta = beta
self.autoencoder_model = self.build_model()
self.autoencoder_model.compile(loss='mse', optimizer=optimizer)