Datasets:
Commit
•
d579ec8
1
Parent(s):
a166f68
Delete loading script
Browse files
vctk.py
DELETED
@@ -1,133 +0,0 @@
|
|
1 |
-
# coding=utf-8
|
2 |
-
# Copyright 2021 The TensorFlow Datasets Authors and the HuggingFace Datasets Authors.
|
3 |
-
#
|
4 |
-
# Licensed under the Apache License, Version 2.0 (the "License");
|
5 |
-
# you may not use this file except in compliance with the License.
|
6 |
-
# You may obtain a copy of the License at
|
7 |
-
#
|
8 |
-
# http://www.apache.org/licenses/LICENSE-2.0
|
9 |
-
#
|
10 |
-
# Unless required by applicable law or agreed to in writing, software
|
11 |
-
# distributed under the License is distributed on an "AS IS" BASIS,
|
12 |
-
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
13 |
-
# See the License for the specific language governing permissions and
|
14 |
-
# limitations under the License.
|
15 |
-
|
16 |
-
# Lint as: python3
|
17 |
-
"""VCTK dataset."""
|
18 |
-
|
19 |
-
|
20 |
-
import os
|
21 |
-
import re
|
22 |
-
|
23 |
-
import datasets
|
24 |
-
from datasets.tasks import AutomaticSpeechRecognition
|
25 |
-
|
26 |
-
|
27 |
-
_CITATION = """\
|
28 |
-
@inproceedings{Veaux2017CSTRVC,
|
29 |
-
title = {CSTR VCTK Corpus: English Multi-speaker Corpus for CSTR Voice Cloning Toolkit},
|
30 |
-
author = {Christophe Veaux and Junichi Yamagishi and Kirsten MacDonald},
|
31 |
-
year = 2017
|
32 |
-
}
|
33 |
-
"""
|
34 |
-
|
35 |
-
_DESCRIPTION = """\
|
36 |
-
The CSTR VCTK Corpus includes speech data uttered by 110 English speakers with various accents.
|
37 |
-
"""
|
38 |
-
|
39 |
-
_URL = "https://datashare.ed.ac.uk/handle/10283/3443"
|
40 |
-
_DL_URL = "https://datashare.is.ed.ac.uk/bitstream/handle/10283/3443/VCTK-Corpus-0.92.zip"
|
41 |
-
|
42 |
-
|
43 |
-
class VCTK(datasets.GeneratorBasedBuilder):
|
44 |
-
"""VCTK dataset."""
|
45 |
-
|
46 |
-
VERSION = datasets.Version("0.9.2")
|
47 |
-
|
48 |
-
BUILDER_CONFIGS = [
|
49 |
-
datasets.BuilderConfig(name="main", version=VERSION, description="VCTK dataset"),
|
50 |
-
]
|
51 |
-
|
52 |
-
def _info(self):
|
53 |
-
return datasets.DatasetInfo(
|
54 |
-
description=_DESCRIPTION,
|
55 |
-
features=datasets.Features(
|
56 |
-
{
|
57 |
-
"speaker_id": datasets.Value("string"),
|
58 |
-
"audio": datasets.features.Audio(sampling_rate=48_000),
|
59 |
-
"file": datasets.Value("string"),
|
60 |
-
"text": datasets.Value("string"),
|
61 |
-
"text_id": datasets.Value("string"),
|
62 |
-
"age": datasets.Value("string"),
|
63 |
-
"gender": datasets.Value("string"),
|
64 |
-
"accent": datasets.Value("string"),
|
65 |
-
"region": datasets.Value("string"),
|
66 |
-
"comment": datasets.Value("string"),
|
67 |
-
}
|
68 |
-
),
|
69 |
-
supervised_keys=("file", "text"),
|
70 |
-
homepage=_URL,
|
71 |
-
citation=_CITATION,
|
72 |
-
task_templates=[AutomaticSpeechRecognition(audio_column="audio", transcription_column="text")],
|
73 |
-
)
|
74 |
-
|
75 |
-
def _split_generators(self, dl_manager):
|
76 |
-
root_path = dl_manager.download_and_extract(_DL_URL)
|
77 |
-
|
78 |
-
return [
|
79 |
-
datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"root_path": root_path}),
|
80 |
-
]
|
81 |
-
|
82 |
-
def _generate_examples(self, root_path):
|
83 |
-
"""Generate examples from the VCTK corpus root path."""
|
84 |
-
|
85 |
-
meta_path = os.path.join(root_path, "speaker-info.txt")
|
86 |
-
txt_root = os.path.join(root_path, "txt")
|
87 |
-
wav_root = os.path.join(root_path, "wav48_silence_trimmed")
|
88 |
-
# NOTE: "comment" is handled separately in logic below
|
89 |
-
fields = ["speaker_id", "age", "gender", "accent", "region"]
|
90 |
-
|
91 |
-
key = 0
|
92 |
-
with open(meta_path, encoding="utf-8") as meta_file:
|
93 |
-
_ = next(iter(meta_file))
|
94 |
-
for line in meta_file:
|
95 |
-
data = {}
|
96 |
-
line = line.strip()
|
97 |
-
search = re.search(r"\(.*\)", line)
|
98 |
-
if search is None:
|
99 |
-
data["comment"] = ""
|
100 |
-
else:
|
101 |
-
start, _ = search.span()
|
102 |
-
data["comment"] = line[start:]
|
103 |
-
line = line[:start]
|
104 |
-
values = line.split()
|
105 |
-
for i, field in enumerate(fields):
|
106 |
-
if field == "region":
|
107 |
-
data[field] = " ".join(values[i:])
|
108 |
-
else:
|
109 |
-
data[field] = values[i] if i < len(values) else ""
|
110 |
-
speaker_id = data["speaker_id"]
|
111 |
-
speaker_txt_path = os.path.join(txt_root, speaker_id)
|
112 |
-
speaker_wav_path = os.path.join(wav_root, speaker_id)
|
113 |
-
# NOTE: p315 does not have text
|
114 |
-
if not os.path.exists(speaker_txt_path):
|
115 |
-
continue
|
116 |
-
for txt_file in sorted(os.listdir(speaker_txt_path)):
|
117 |
-
filename, _ = os.path.splitext(txt_file)
|
118 |
-
_, text_id = filename.split("_")
|
119 |
-
for i in [1, 2]:
|
120 |
-
wav_file = os.path.join(speaker_wav_path, f"{filename}_mic{i}.flac")
|
121 |
-
# NOTE: p280 does not have mic2 files
|
122 |
-
if not os.path.exists(wav_file):
|
123 |
-
continue
|
124 |
-
with open(os.path.join(speaker_txt_path, txt_file), encoding="utf-8") as text_file:
|
125 |
-
text = text_file.readline().strip()
|
126 |
-
more_data = {
|
127 |
-
"file": wav_file,
|
128 |
-
"audio": wav_file,
|
129 |
-
"text": text,
|
130 |
-
"text_id": text_id,
|
131 |
-
}
|
132 |
-
yield key, {**data, **more_data}
|
133 |
-
key += 1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|