""" Decoder """
# image -> word
decoder_trainer = DecoderTrainer(state_encoder,
state_decoder,
decoder_gan,
training_epochs=constants.TRAINING_EPOCHS,
batch_size=constants.BATCH_SIZE)
"""
Start training
"""
text_displayer = SampleTextDisplayer()
diagram_displayer = SampleDiagramDisplayer()
image_displayer = SampleImageDisplayer(row=constants.DISPLAY_ROW,
column=constants.DISPLAY_COLUMN,
cmap='gray')
seq2seq_loss = []
seq2seq_accuracy = []
encoder_discriminator_loss = []
encoder_discriminator_accuracy = []
encoder_generator_loss = []
encoder_generator_accuracy = []
decoder_loss = []
decoder_accuracy = []
# Create decoder trainer
decoder_trainer = DecoderTrainer(encoder_generator,
decoder_generator,
decoder_gan,
training_epochs=constants.TRAINING_EPOCHS,
batch_size=constants.BATCH_SIZE)
"""
Start training
"""
image_displayer = SampleImageDisplayer(row=constants.DISPLAY_ROW,
column=constants.DISPLAY_COLUMN,
cmap='gray')
diagram_displayer = SampleDiagramDisplayer()
confusion_displayer = SampleConfusionMatrixDisplayer()
report_displayer = SampleReportDisplayer()
encoder_discriminator_loss = []
encoder_discriminator_accuracy = []
encoder_generator_loss = []
encoder_generator_accuracy = []
decoder_loss = []
decoder_accuracy = []
class_loss = []
class_accuracy = []
print('x_test: {}'.format(x_test.shape))
print('y_train: {}'.format(y_train.shape))
print('y_test: {}'.format(y_test.shape))
# Create classifier
classifier_creator = ClassifierModelCreator(constants.INPUT_SHAPE, 10,
model_name)
classifier = classifier_creator.create_model()
# Train classifier
trainer = ClassifierTrainer(classifier, constants.TRAINING_EPOCHS,
constants.BATCH_SIZE)
accuracy, val_accuracy = trainer.train(x_train, y_train, x_test, y_test)
# Plot history
diagram_displayer = SampleDiagramDisplayer()
diagram_displayer.display_samples(
name='Classifier Fashion Accuracy',
samples=accuracy,
should_display_directly=should_display_directly,
should_save_to_file=should_save_to_file)
diagram_displayer.display_samples(
name='Classifier Fashion Validation Accuracy',
samples=val_accuracy,
should_display_directly=should_display_directly,
should_save_to_file=should_save_to_file)
# Save classifier model for future use
model_path = 'model/{}.h5'.format(model_name)
classifier.save(model_path)