mirror of
https://github.com/Nighthawk42/MioTTS.git
synced 2026-08-30 09:42:27 +00:00
105 lines
3.2 KiB
Python
105 lines
3.2 KiB
Python
import argparse
|
|
import os
|
|
from typing import Tuple
|
|
|
|
|
|
def _validate_dataset(root: str, metadata_name: str, wavs_subdir: str) -> Tuple[int, int, int]:
|
|
valid_lines = 0
|
|
invalid_lines = 0
|
|
missing_wavs = 0
|
|
|
|
metadata_path = os.path.join(root, metadata_name)
|
|
if not os.path.isfile(metadata_path):
|
|
msg = f"metadata file not found: {metadata_path}"
|
|
raise FileNotFoundError(msg)
|
|
|
|
wavs_root = os.path.join(root, wavs_subdir)
|
|
if not os.path.isdir(wavs_root):
|
|
msg = f"wavs directory not found: {wavs_root}"
|
|
raise FileNotFoundError(msg)
|
|
|
|
with open(metadata_path, encoding="utf-8-sig") as f:
|
|
for idx, raw_line in enumerate(f, start=1):
|
|
line = raw_line.strip()
|
|
if not line:
|
|
continue
|
|
|
|
parts = line.split("|")
|
|
if len(parts) != 3:
|
|
invalid_lines += 1
|
|
print(f"[line {idx}] invalid field count ({len(parts)} != 3): {line}")
|
|
continue
|
|
|
|
file_id, normalized_text, original_text = parts
|
|
file_id = file_id.strip()
|
|
normalized_text = normalized_text.strip()
|
|
original_text = original_text.strip()
|
|
|
|
if not file_id or any(sep in file_id for sep in ("/", "\\")):
|
|
invalid_lines += 1
|
|
print(f"[line {idx}] invalid file_id: {file_id!r}")
|
|
continue
|
|
|
|
if not normalized_text:
|
|
invalid_lines += 1
|
|
print(f"[line {idx}] empty normalized_text")
|
|
continue
|
|
|
|
if not original_text:
|
|
invalid_lines += 1
|
|
print(f"[line {idx}] empty original_text")
|
|
continue
|
|
|
|
wav_path = os.path.join(wavs_root, f"{file_id}.wav")
|
|
if not os.path.isfile(wav_path):
|
|
missing_wavs += 1
|
|
print(f"[line {idx}] missing wav file: {wav_path}")
|
|
|
|
valid_lines += 1
|
|
|
|
return valid_lines, invalid_lines, missing_wavs
|
|
|
|
|
|
def main() -> None:
|
|
parser = argparse.ArgumentParser(description="Validate an LJSpeech-style dataset folder.")
|
|
parser.add_argument(
|
|
"--root",
|
|
required=True,
|
|
help="Path to LJSpeech dataset root containing wavs/ and metadata.csv.",
|
|
)
|
|
parser.add_argument(
|
|
"--metadata",
|
|
default="metadata.csv",
|
|
help="Metadata filename relative to root (default: metadata.csv).",
|
|
)
|
|
parser.add_argument(
|
|
"--wavs-subdir",
|
|
default="wavs",
|
|
help="Subdirectory under root containing wav files (default: wavs).",
|
|
)
|
|
args = parser.parse_args()
|
|
|
|
root = os.path.abspath(args.root)
|
|
if not os.path.isdir(root):
|
|
msg = f"dataset root is not a directory: {root}"
|
|
raise NotADirectoryError(msg)
|
|
|
|
print(f"Validating LJSpeech dataset in: {root}")
|
|
print(f" metadata: {args.metadata}")
|
|
print(f" wavs subdir: {args.wavs_subdir}")
|
|
|
|
valid, invalid, missing = _validate_dataset(root, args.metadata, args.wavs_subdir)
|
|
|
|
print()
|
|
print(f"Valid lines : {valid}")
|
|
print(f"Invalid lines : {invalid}")
|
|
print(f"Missing wav files: {missing}")
|
|
|
|
if invalid or missing:
|
|
raise SystemExit(1)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
main()
|
|
|