os.path.join(wavs_path, fn + ".wav"), wav, hp.sr)
chapter += 1
except FileNotFoundError:
break
_convert_mp3_to_wav('01_Genesis', 1)
_convert_mp3_to_wav('02_Exodus', 2)
_convert_mp3_to_wav('03_Leviticus', 3)
metadata_csv.close()
print("total audio duration: %ss" %
(time.strftime('%H:%M:%S', time.gmtime(total_duration_s))))
# pre process
print("pre processing...")
mb_speech = MBSpeech([])
preprocess(dataset_path, mb_speech)
else:
if not args.dataset:
print("You must specify a dataset")
sys.exit(0)
dataset_name = args.dataset
datasets_path = os.path.join(os.path.dirname(os.path.realpath(__file__)),
'datasets')
dataset_path = os.path.join(datasets_path, dataset_name)
if not os.path.isdir(dataset_path):
print("%s does not exist" % dataset_path)
print("You should put your unprocessed dataset inside %s" %
datasets_path)
sys.exit(0)
import argparse
from tqdm import *
import numpy as np
from torch.utils.data import ConcatDataset
from datasets import Compose, LoadAudio, ComputeMagSpectrogram
parser = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument("--dataset",
choices=['librispeech', 'mbspeech', 'bolorspeech', 'backgroundsounds'],
default='bolorspeech', help='dataset name')
args = parser.parse_args()
if args.dataset == 'mbspeech':
from datasets.mb_speech import MBSpeech
dataset = MBSpeech()
elif args.dataset == 'librispeech':
from datasets.libri_speech import LibriSpeech
dataset = ConcatDataset([
LibriSpeech(name='train-clean-100'),
LibriSpeech(name='train-clean-360'),
LibriSpeech(name='train-other-500'),
LibriSpeech(name='dev-clean',)
])
elif args.dataset == 'backgroundsounds':
from datasets.background_sounds import BackgroundSounds
dataset = BackgroundSounds(is_random=False)
elif args.dataset == 'bolorspeech':
from datasets.bolor_speech import BolorSpeech
dataset = ConcatDataset([
BolorSpeech(name='train'),