import matplotlib.pyplot as plt from copy import deepcopy from gantools import blocks from audioinpainting.load import load_audio_dataset from audioinpainting.model_basic import InpaintingGAN # # Parameters downscale = 2 # # Data handling # Load the data start = time.time() # dataset = data.load.load_audio_dataset(scaling=downscale) dataset = load_audio_dataset(scaling=downscale, type='solo', spix=1024 * 16, augmentation=True) print('Number of samples: {}'.format(dataset.N)) # ============================================================================= # # The dataset can return an iterator. # it = dataset.iter(10) # print(next(it).shape) # del it # # # Get all the data # X = dataset.get_all_data().flatten() # # plt.hist(X, 100) # print('min: {}'.format(np.min(X)))
uu = 0 + aa spix = 1024 * 52 signal_length = 1024 * 52 signal_split = [1024 * 18, 1024 * 6, 1024 * 4, 1024 * 6, 1024 * 18] elif model == 'basic': from audioinpainting.model_basic import InpaintingGAN uu = 50 + bb spix = 1024 * 52 signal_length = 1024 * 52 signal_split = [1024 * 24, 1024 * 4, 1024 * 24] else: raise ValueError( 'Incorrect model; choose either "extend" or "basic"') dataset = load_audio_dataset(scaling=downscale, type=type, spix=spix, augmentation=True) # Check whether number of generated samples is consistent with total number of samples if N_f > dataset.N: N_f = dataset.N print('Number of samples: {}'.format(dataset.N)) # ## Parameters bn = False md = 64 params_discriminator = dict() params_discriminator['stride'] = [4, 4, 4, 4, 4] params_discriminator['nfilter'] = [md, 2 * md, 4 * md, 8 * md, 16 * md] params_discriminator['shape'] = [[25], [25], [25], [25], [25]]
import matplotlib.pyplot as plt from copy import deepcopy from gantools import blocks from audioinpainting.load import load_audio_dataset from audioinpainting.model_extend import InpaintingGAN # # Parameters downscale = 2 # # Data handling # Load the data start = time.time() # dataset = data.load.load_audio_dataset(scaling=downscale) dataset = load_audio_dataset(scaling=downscale, type='piano', spix=1024*52, augmentation=True) print('Number of samples: {}'.format(dataset.N)) # ============================================================================= # # The dataset can return an iterator. # it = dataset.iter(10) # print(next(it).shape) # del it # # # Get all the data # X = dataset.get_all_data().flatten() # # plt.hist(X, 100) # print('min: {}'.format(np.min(X)))