width_shift_range=0.2,
height_shift_range=0.2,
horizontal_flip=True)
datagen.fit(X_train)
y_train_matrix = to_categorical(y_train, len(labeltonumber))
y_val_matrix = to_categorical(y_val, len(labeltonumber))
print('[INFO] start training...')
if modus == 'train':
if worker == 'single':
## use validation fold for validation
model.fit_generator(datagen.flow(X_train,
y_train_matrix,
batch_size=BATCHSIZE),
validation_data=[X_val, y_val_matrix],
epochs=EPOCHS,
steps_per_epoch=STEPS_PER_EPOCH,
verbose=1,
callbacks=[checkpoint, lr_scheduler, csv_logger])
elif worker == 'parallel':
parallel_model.fit_generator(
datagen.flow(X_train, y_train_matrix, batch_size=BATCHSIZE),
validation_data=[X_val, y_val_matrix],
epochs=EPOCHS,
steps_per_epoch=STEPS_PER_EPOCH,
verbose=1,
callbacks=[checkpoint, lr_scheduler, csv_logger])
###################################### end version basic and balance ###################################################
####################################### run on test set ################################################################
class_mode='categorical')
print(train_generator.class_indices)
validation_generator = datagen.flow_from_directory(validation_data_dir,
classes=cls_folder,
target_size=(img_height,
img_width),
batch_size=batch_size,
shuffle=True,
class_mode='categorical')
filepath = 'ep{epoch:03d}-val_loss{val_loss:.4f}-val_acc{val_acc:.4f}.h5'
checkpoint = ModelCheckpoint(os.path.join(saving_path, filepath),
monitor='val_loss',
verbose=1,
save_best_only=False,
save_weights_only=True,
mode='min',
period=1)
board = TensorBoard(log_dir='../../tmp/log',
histogram_freq=0,
embeddings_freq=0,
write_images=True)
early_stoping = EarlyStopping(monitor='acc', patience=3)
model.fit_generator(train_generator,
steps_per_epoch=nb_train_samples // batch_size,
epochs=epochs,
validation_data=validation_generator,
validation_steps=nb_validation_samples // batch_size,
callbacks=[board, checkpoint, early_stoping])
