train_cnn_augmented_data.py File Reference

Go to the source code of this file.

Namespaces
	train_cnn_augmented_data

Functions
def	train_cnn_augmented_data.save_model (model, name)
def	train_cnn_augmented_data.generate_data_generator (generator, X, Y1, Y2, b)

Variables
	train_cnn_augmented_data.parser = argparse.ArgumentParser(description='Run CNN training on patches with a few different hyperparameter sets.')
	train_cnn_augmented_data.help
	train_cnn_augmented_data.default
	train_cnn_augmented_data.args = parser.parse_args()
	train_cnn_augmented_data.config = read_config(args.config)
	configuration ############################# More...
	train_cnn_augmented_data.CNN_INPUT_DIR = config['training_on_patches']['input_dir']
	train_cnn_augmented_data.PATCH_SIZE_W
	train_cnn_augmented_data.PATCH_SIZE_D
	train_cnn_augmented_data.img_rows
	train_cnn_augmented_data.img_cols
	train_cnn_augmented_data.batch_size = config['training_on_patches']['batch_size']
	train_cnn_augmented_data.nb_classes = config['training_on_patches']['nb_classes']
	train_cnn_augmented_data.nb_epoch = config['training_on_patches']['nb_epoch']
int	train_cnn_augmented_data.nb_pool = 2
string	train_cnn_augmented_data.cfg_name = 'sgd_lorate'
int	train_cnn_augmented_data.nb_filters1 = 48
int	train_cnn_augmented_data.nb_conv1 = 5
string	train_cnn_augmented_data.convactfn1 = 'relu'
bool	train_cnn_augmented_data.maxpool = False
int	train_cnn_augmented_data.nb_filters2 = 0
int	train_cnn_augmented_data.nb_conv2 = 7
string	train_cnn_augmented_data.convactfn2 = 'relu'
float	train_cnn_augmented_data.drop1 = 0.2
int	train_cnn_augmented_data.densesize1 = 128
string	train_cnn_augmented_data.actfn1 = 'relu'
int	train_cnn_augmented_data.densesize2 = 32
string	train_cnn_augmented_data.actfn2 = 'relu'
float	train_cnn_augmented_data.drop2 = 0.2
	train_cnn_augmented_data.main_input = Input(shape=(img_rows, img_cols, 1), name='main_input')
	CNN definition ############################. More...
	train_cnn_augmented_data.x
	train_cnn_augmented_data.em_trk_none = Dense(3, activation='softmax', name='em_trk_none_netout')(x)
	train_cnn_augmented_data.michel = Dense(1, activation='sigmoid', name='michel_netout')(x)
	train_cnn_augmented_data.sgd = SGD(lr=0.01, decay=1e-5, momentum=0.9, nesterov=True)
	train_cnn_augmented_data.model = Model(inputs=[main_input], outputs=[em_trk_none, michel])
	train_cnn_augmented_data.optimizer
	train_cnn_augmented_data.loss
	train_cnn_augmented_data.loss_weights
	train_cnn_augmented_data.n_training = count_events(CNN_INPUT_DIR, 'training')
	read data sets ############################ More...
	train_cnn_augmented_data.X_train = np.zeros((n_training, PATCH_SIZE_W, PATCH_SIZE_D, 1), dtype=np.float32)
	train_cnn_augmented_data.EmTrkNone_train = np.zeros((n_training, 3), dtype=np.int32)
	train_cnn_augmented_data.Michel_train = np.zeros((n_training, 1), dtype=np.int32)
int	train_cnn_augmented_data.ntot = 0
list	train_cnn_augmented_data.subdirs = [f for f in os.listdir(CNN_INPUT_DIR) if 'training' in f]
list	train_cnn_augmented_data.filesX = [f for f in os.listdir(CNN_INPUT_DIR + '/' + dirname) if '_x.npy' in f]
	train_cnn_augmented_data.fnameY = fnameX.replace('_x.npy', '_y.npy')
	train_cnn_augmented_data.dataX = np.load(CNN_INPUT_DIR + '/' + dirname + '/' + fnameX)
	train_cnn_augmented_data.dataY = np.load(CNN_INPUT_DIR + '/' + dirname + '/' + fnameY)
	train_cnn_augmented_data.n = dataY.shape[0]
	train_cnn_augmented_data.n_testing = count_events(CNN_INPUT_DIR, 'testing')
	train_cnn_augmented_data.X_test = np.zeros((n_testing, PATCH_SIZE_W, PATCH_SIZE_D, 1), dtype=np.float32)
	train_cnn_augmented_data.EmTrkNone_test = np.zeros((n_testing, 3), dtype=np.int32)
	train_cnn_augmented_data.Michel_test = np.zeros((n_testing, 1), dtype=np.int32)
	train_cnn_augmented_data.datagen
	training ############################### More...
	train_cnn_augmented_data.h
	train_cnn_augmented_data.score