Spaces:
Runtime error
Runtime error
File size: 5,110 Bytes
5019931 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 |
import argparse
import os
from typing import NoReturn
import librosa
import numpy as np
import soundfile
from bytesep.dataset_creation.pack_audios_to_hdf5s.instruments_solo import (
read_csv as read_instruments_solo_csv,
)
from bytesep.dataset_creation.pack_audios_to_hdf5s.maestro import (
read_csv as read_maestro_csv,
)
from bytesep.utils import load_random_segment
def create_evaluation(args) -> NoReturn:
r"""Random mix and write out audios for evaluation.
Args:
piano_dataset_dir: str, the directory of the piano dataset
symphony_dataset_dir: str, the directory of the symphony dataset
evaluation_audios_dir: str, the directory to write out randomly selected and mixed audio segments
sample_rate: int
channels: int, e.g., 1 | 2
evaluation_segments_num: int
mono: bool
Returns:
NoReturn
"""
# arguments & parameters
piano_dataset_dir = args.piano_dataset_dir
symphony_dataset_dir = args.symphony_dataset_dir
evaluation_audios_dir = args.evaluation_audios_dir
sample_rate = args.sample_rate
channels = args.channels
evaluation_segments_num = args.evaluation_segments_num
mono = True if channels == 1 else False
split = 'test'
segment_seconds = 10.0
random_state = np.random.RandomState(1234)
piano_meta_csv = os.path.join(piano_dataset_dir, 'maestro-v2.0.0.csv')
piano_names_dict = read_maestro_csv(piano_meta_csv)
piano_audio_names = piano_names_dict[split]
symphony_meta_csv = os.path.join(symphony_dataset_dir, 'validation.csv')
symphony_names_dict = read_instruments_solo_csv(symphony_meta_csv)
symphony_audio_names = symphony_names_dict[split]
for source_type in ['piano', 'symphony', 'mixture']:
output_dir = os.path.join(evaluation_audios_dir, split, source_type)
os.makedirs(output_dir, exist_ok=True)
for n in range(evaluation_segments_num):
print('{} / {}'.format(n, evaluation_segments_num))
# Randomly select and write out a clean piano segment.
piano_audio_name = random_state.choice(piano_audio_names)
piano_audio_path = os.path.join(piano_dataset_dir, piano_audio_name)
piano_audio = load_random_segment(
audio_path=piano_audio_path,
random_state=random_state,
segment_seconds=segment_seconds,
mono=mono,
sample_rate=sample_rate,
)
output_piano_path = os.path.join(
evaluation_audios_dir, split, 'piano', '{:04d}.wav'.format(n)
)
soundfile.write(
file=output_piano_path, data=piano_audio.T, samplerate=sample_rate
)
print("Write out to {}".format(output_piano_path))
# Randomly select and write out a clean symphony segment.
symphony_audio_name = random_state.choice(symphony_audio_names)
symphony_audio_path = os.path.join(
symphony_dataset_dir, "mp3s", symphony_audio_name
)
symphony_audio = load_random_segment(
audio_path=symphony_audio_path,
random_state=random_state,
segment_seconds=segment_seconds,
mono=mono,
sample_rate=sample_rate,
)
output_symphony_path = os.path.join(
evaluation_audios_dir, split, 'symphony', '{:04d}.wav'.format(n)
)
soundfile.write(
file=output_symphony_path, data=symphony_audio.T, samplerate=sample_rate
)
print("Write out to {}".format(output_symphony_path))
# Mix piano and symphony segments and write out a mixture segment.
mixture_audio = symphony_audio + piano_audio
output_mixture_path = os.path.join(
evaluation_audios_dir, split, 'mixture', '{:04d}.wav'.format(n)
)
soundfile.write(
file=output_mixture_path, data=mixture_audio.T, samplerate=sample_rate
)
print("Write out to {}".format(output_mixture_path))
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
"--piano_dataset_dir",
type=str,
required=True,
help="The directory of the piano dataset.",
)
parser.add_argument(
"--symphony_dataset_dir",
type=str,
required=True,
help="The directory of the symphony dataset.",
)
parser.add_argument(
"--evaluation_audios_dir",
type=str,
required=True,
help="The directory to write out randomly selected and mixed audio segments.",
)
parser.add_argument(
"--sample_rate",
type=int,
required=True,
help="Sample rate.",
)
parser.add_argument(
"--channels",
type=int,
required=True,
help="Audio channels, e.g, 1 or 2.",
)
parser.add_argument(
"--evaluation_segments_num",
type=int,
required=True,
help="The number of segments to create for evaluation.",
)
# Parse arguments.
args = parser.parse_args()
create_evaluation(args)
|