Complete_Skips.py

#!/usr/bin/env ipython3
# initialization
# ---------------------------------------------------------------------------------------------------------------------------------------------------

length_file = 2.3
SNR_matrix = [-10,0,10]
option = 'most'
path_s = '/media/aneumann/Harddisk/Bachelorarbeit/TIMIT/timit_16kHz_wav'
path_m = '/media/aneumann/Harddisk/Bachelorarbeit/MedleyDB/Audio'
path_save = '/media/aneumann/Harddisk/Bachelorarbeit/BATest'
# path_s = '/ba/TIMIT/timit_16kHz_wav_concatenated/train'
# path_m = '/ba/MedleyDB/Audio'
# path_save = '/ba/BATest'
testname = '/test_skip_15'
pattern = "*.wav"
split = 0.8

# Abspeichern
path_save_augmentation = path_save + '/Gemischte_Signale'

# Preprocessing
# ---------------------------------------------------------------------------------------------------------------------------------------------------

path_load = path_save_augmentation
# path_save_preprocessing = path_save + '/test_framelength_5/Preprocessing'
path_save_preprocessing = path_save + '/test_framelength/Preprocessing'
option = 'music'
framelength = 15

# Neuronales Netz trainieren
# ---------------------------------------------------------------------------------------------------------------------------------------------------

path_train = path_save_preprocessing + '/Train'
path_valid = path_save_preprocessing + '/Validation'
batch_size = 20
epochs = 5
reduction_divisor = 10
path_save_model = path_save + testname + '/cnn_skip_15_100epochs.h5'
path_weights_model = path_save + testname + '/cnn_skip_15_100epochs_weights'
plot_name = 'test'
weight = 'weights-92-450.2259.hdf5'

# from Skip_Autoencoder import skip_autoencoder
# print('Training CNN Autoencoder ...')
# skip_autoencoder(path_train, path_valid, batch_size, epochs, path_save_model, path_weights_model, weight, plot_name, option, reduction_divisor)

# Prediction
# ---------------------------------------------------------------------------------------------------------------------------------------------------

path_data = path_save_preprocessing
path_save_prediction = path_save + testname + '/Predict_skip_15_100epochs'
option_dir = 'test' # or 'train' or 'valid'
# option2 = framelength
# weight_path = '/media/aneumann/Harddisk/Bachelorarbeit/BATest/test_skip_15/cnn_skip_15_100epochs_weights/weights-90-366.7344.hdf5'
import os
weight_paths = os.listdir(path_weights_model)
weight_paths = sorted(weight_paths,key=lambda x: (int(x.split(".")[0][-4:])))
weight_path = path_weights_model + '/' + weight_paths[-1]
weight_path = '/media/aneumann/Harddisk/Bachelorarbeit/BATest/test_skip_15/cnn_skip_15_100epochs_weights/weights-90-366.7344.hdf5'

from Prediction.Prediction_skips import skips_predict

# print('Predicting: CNN ...')
# skips_predict(path_save_model, path_data, path_save_prediction, option_dir, option, weight_path)

# Auswertung
# ---------------------------------------------------------------------------------------------------------------------------------------------------

option_audio = 'noaudiosave'
path_files = path_save_augmentation
path_predicted = path_save_prediction
path_save_ausw = path_save + testname + '/Auswertung_skip_15_100epochs'
K_sub = 4
I_sub = 4

from Auswertung.Auswertung import auswertung_framelength

print('Auswertung CNN ...')
auswertung_framelength(path_files, path_predicted, path_save_ausw, K_sub, I_sub, option ,option_audio, option_dir)