def experiment_4():
config = get_experiment_4_config()
config.logger.start()
loader = ImageDataLoader(config)
trainer = HyperTrainer(config=config, loader=loader)
trainer.start_tunning()
config.logger.end()
def experiment_0():
config = get_experiment_0_config()
config.logger.start()
loader = ImageDataLoader(config)
input_shape = config.input_shape
logging.info(f"input_shape:{input_shape}")
number_of_classes = len(loader.labbel_mapper.classes_name)
logging.info(f"number_of_classes:{number_of_classes}")
trainer = Trainer(config=config, loader=loader)
trainer.start(model=get_mobilenet(
hidden_activation=config.hidden_activation,
output_activation=config.output_activation,
input_shape=input_shape,
number_of_classes=number_of_classes),
run_id="mobilenet_model")
trainer.start(model=get_cnn(hidden_activation=config.hidden_activation,
output_activation=config.output_activation,
input_shape=input_shape,
number_of_classes=number_of_classes),
run_id="cnn_model")
trainer = HyperTrainer(config=config, loader=loader)
trainer.start_tunning()
config.logger.end()
def experiment_1(trainable: bool):
if trainable:
config = get_experiment_1_config(
experiment_name='experiment_1_trainable_true')
else:
config = get_experiment_1_config(
experiment_name='experiment_1_trainable_false')
config.logger.start()
loader = ImageDataLoader(config)
input_shape = config.input_shape
logging.info(f"input_shape:{input_shape}")
number_of_classes = len(loader.labbel_mapper.classes_name)
logging.info(f"number_of_classes:{number_of_classes}")
trainer = Trainer(config=config, loader=loader)
trainer.start(model=get_mobilenet(
hidden_activation=config.hidden_activation,
output_activation=config.output_activation,
input_shape=input_shape,
number_of_classes=number_of_classes,
trainable=trainable),
run_id="mobilenet_model")
trainer.start(model=get_vgg16(hidden_activation=config.hidden_activation,
output_activation=config.output_activation,
input_shape=input_shape,
number_of_classes=number_of_classes,
trainable=trainable),
run_id="vgg16_model")
trainer.start(model=get_densenet121(
hidden_activation=config.hidden_activation,
output_activation=config.output_activation,
input_shape=input_shape,
number_of_classes=number_of_classes,
trainable=trainable),
run_id="densenet121_model")
trainer.start(model=get_resnet50(
hidden_activation=config.hidden_activation,
output_activation=config.output_activation,
input_shape=input_shape,
number_of_classes=number_of_classes,
trainable=trainable),
run_id="resnet50_model")
trainer.start(model=get_cnn(hidden_activation=config.hidden_activation,
output_activation=config.output_activation,
input_shape=input_shape,
number_of_classes=number_of_classes),
run_id="cnn_model")
config.logger.end()
def experiment_3_evaluate():
config = get_experiment_3_config()
config.logger.start()
loader = ImageDataLoader(config)
input_shape = config.input_shape
logging.info(f"input_shape:{input_shape}")
number_of_classes = len(loader.labbel_mapper.classes_name)
logging.info(f"number_of_classes:{number_of_classes}")
trainer = Trainer(config=config, loader=loader)
trainer.evaluate(run_id='cnn_model',
test_data=loader.test_data,
steps=loader.test_data_info.count)
config.logger.end()
def main():
config = get_experiment_1_config()
logger = Logger(config)
logger.start()
loader = ImageDataLoader(config)
input_shape = config.input_shape
logging.info(f"input_shape:{input_shape}")
number_of_classes = len(loader.labbel_mapper.labels)
logging.info(f"number_of_classes:{number_of_classes}")
trainer = Trainer(config=config, loader=loader, logger=logger)
trainer.start(model=get_cnn(hidden_activation=config.hidden_activation,
output_activation=config.output_activation,
input_shape=input_shape,
number_of_classes=number_of_classes))
logger.end()
def get_data_loader(self):
return ImageDataLoader(self.config)