MycroftAI
/
mimic2
mirror of https://github.com/MycroftAI/mimic2.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

Replace Amy with Danny data

danny
Keith Ito 2018-04-23 10:28:08 -07:00
parent c7efb3c208
commit 59f51f7197
2 changed files with 12 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
datasets/amy.py → datasets/danny.py
View File

@ -40,6 +40,9 @@ def _process_utterance(out_dir, prompt_id, wav_path, text):
wav = wav[margin:-margin] wav = wav[margin:-margin]
wav, _ = librosa.effects.trim(wav, top_db=40, frame_length=1024, hop_length=256) wav, _ = librosa.effects.trim(wav, top_db=40, frame_length=1024, hop_length=256)
# Normalize
wav /= max(0.01, np.max(np.abs(wav)))
# Compute the linear-scale spectrogram from the wav: # Compute the linear-scale spectrogram from the wav:
spectrogram = audio.spectrogram(wav).astype(np.float32) spectrogram = audio.spectrogram(wav).astype(np.float32)
n_frames = spectrogram.shape[1] n_frames = spectrogram.shape[1]
@ -48,8 +51,8 @@ def _process_utterance(out_dir, prompt_id, wav_path, text):
mel_spectrogram = audio.melspectrogram(wav).astype(np.float32) mel_spectrogram = audio.melspectrogram(wav).astype(np.float32)
# Write the spectrograms to disk: # Write the spectrograms to disk:
spectrogram_filename = 'amy-spec-%s.npy' % prompt_id spectrogram_filename = 'danny-spec-%s.npy' % prompt_id
mel_filename = 'amy-mel-%s.npy' % prompt_id mel_filename = 'danny-mel-%s.npy' % prompt_id
np.save(os.path.join(out_dir, spectrogram_filename), spectrogram.T, allow_pickle=False) np.save(os.path.join(out_dir, spectrogram_filename), spectrogram.T, allow_pickle=False)
np.save(os.path.join(out_dir, mel_filename), mel_spectrogram.T, allow_pickle=False) np.save(os.path.join(out_dir, mel_filename), mel_spectrogram.T, allow_pickle=False)

14
preprocess.py
View File

@ -2,7 +2,7 @@ import argparse
import os import os
from multiprocessing import cpu_count from multiprocessing import cpu_count
from tqdm import tqdm from tqdm import tqdm
from datasets import amy, blizzard, ljspeech from datasets import danny, blizzard, ljspeech
from hparams import hparams from hparams import hparams
@ -22,11 +22,11 @@ def preprocess_ljspeech(args):
write_metadata(metadata, out_dir) write_metadata(metadata, out_dir)
def preprocess_amy(args): def preprocess_danny(args):
in_dir = os.path.join(args.base_dir, 'amy') in_dir = os.path.join(args.base_dir, 'danny')
out_dir = os.path.join(args.base_dir, args.output) out_dir = os.path.join(args.base_dir, args.output)
os.makedirs(out_dir, exist_ok=True) os.makedirs(out_dir, exist_ok=True)
metadata = amy.build_from_path(in_dir, out_dir, args.num_workers, tqdm=tqdm) metadata = danny.build_from_path(in_dir, out_dir, args.num_workers, tqdm=tqdm)
write_metadata(metadata, out_dir) write_metadata(metadata, out_dir)
@ -45,11 +45,11 @@ def main():
parser = argparse.ArgumentParser() parser = argparse.ArgumentParser()
parser.add_argument('--base_dir', default=os.path.expanduser('~/tacotron')) parser.add_argument('--base_dir', default=os.path.expanduser('~/tacotron'))
parser.add_argument('--output', default='training') parser.add_argument('--output', default='training')
parser.add_argument('--dataset', required=True, choices=['amy', 'blizzard', 'ljspeech']) parser.add_argument('--dataset', required=True, choices=['danny', 'blizzard', 'ljspeech'])
parser.add_argument('--num_workers', type=int, default=cpu_count()) parser.add_argument('--num_workers', type=int, default=cpu_count())
args = parser.parse_args() args = parser.parse_args()
if args.dataset == 'amy': if args.dataset == 'danny':
preprocess_amy(args) preprocess_danny(args)
elif args.dataset == 'blizzard': elif args.dataset == 'blizzard':
preprocess_blizzard(args) preprocess_blizzard(args)
elif args.dataset == 'ljspeech': elif args.dataset == 'ljspeech':