unofficial mirror of Vietnam-Celeb dataset

official announcement:

• https://www.isca-archive.org/interspeech_2023/pham23b_interspeech.html
• https://github.com/Vietnam-Celeb/Vietnam-Celeb
• https://huggingface.co/datasets/hustep-lab/Vietnam-Celeb

official download:

• Part 0: https://drive.google.com/file/d/1pMuT3DFzSwib7SVcRS8VkDwPuLTsemSG/view?usp=share_link
• Part 1: https://drive.google.com/file/d/1xayHt2HRqE1aJ4HvtUT40_9XlgvfDfRY/view?usp=share_link
• Part 2: https://drive.google.com/file/d/1MIlM78EbN_J9cApkNes_2_BrFrf8XwMc/view?usp=share_link
• Part 3: https://drive.google.com/file/d/1h6Na58DC03p-502B9QpC5Z_FadUwAdNA/view?usp=share_link

DOI: 10.21437/Interspeech.2023-1989

187h, 87k samples, no transcription

additional data:

• speaker metadata: see speaker-metadata.csv
• easy part of test set: see vietnam-celeb-e.csv
• hard part of test set: see vietnam-celeb-h.csv

easy/hard test is alike to VoxCeleb dataset

pre-process: see my code: https://github.com/phineas-pta/fine-tune-whisper-vi/blob/main/misc/viet-celeb.py

usage with HuggingFace:

# pip install -q "datasets[audio]"
from datasets import load_dataset
from torch.utils.data import DataLoader

dataset = load_dataset("doof-ferb/Vietnam-Celeb", split="train", streaming=True)
dataset.set_format(type="torch", columns=["audio"])
dataloader = DataLoader(dataset, batch_size=4)