Datasetpublic

SpeechAntiSpoofingBenchmarks/ASVspoof2021_LA

ASVspoof 2021 LA Benchmark-ready packaging of the Logical Access (LA) evaluation partition from ASVspoof 2021 for speech anti-spoofing and synthetic / deepfake voice detection. Overview This dataset contains the LA evaluation subset of the ASVspoof 2021 challenge. The task is binary classification: bonafide (genuine human speech) vs. spoof (synthetic, converted, or replayed speech). The LA partition adds realistic telephony codec / transmission conditions on top… See the full description on the dataset page: https://huggingface.co/datasets/SpeechAntiSpoofingBenchmarks/ASVspoof2021_LA.

sourceHugging Faceodc-byupdated 3mo agoView on Hugging Face

0likes1.6kdownloads

Dataset Card

ASVspoof 2021 LA

Benchmark-ready packaging of the Logical Access (LA) evaluation partition from ASVspoof 2021 for speech anti-spoofing and synthetic / deepfake voice detection.

Overview

This dataset contains the LA evaluation subset of the ASVspoof 2021 challenge. The task is binary classification: bonafide (genuine human speech) vs. spoof (synthetic, converted, or replayed speech). The LA partition adds realistic telephony codec / transmission conditions on top of the 2019 LA attacks. The original dataset is available at https://www.asvspoof.org/index2021.html.

License & redistribution

This dataset is redistributed under the Open Data Commons Attribution License (ODC-By). See LICENSE.txt for the full text.

The labels and evaluation protocol are unmodified. The audio was transcoded to canonical 16 kHz mono FLAC: a large fraction of the source FLAC files use an encoding the standard libsndfile/soundfile decoder cannot read (and therefore the Hugging Face datasets audio decoder cannot read), so each clip was decoded and re-encoded to a clean, universally decodable FLAC. The PCM samples are preserved bit-exactly; the sampling rate (16 kHz) is unchanged.

Schema

Column	Type	Description
`path`	`string`	Stable archive-relative path (e.g. `LA_E_9332881.flac`), unique within dataset
`audio`	`Audio(16000)`	Audio waveform, 16 kHz mono
`label`	`ClassLabel`	`"bonafide"` (index 0) or `"spoof"` (index 1)
`notes`	`string`	JSON with `utterance_id`, `speaker_id`, `codec`, `transmission`, `attack_id`, `trim`, `phase`

notes example:

json

{"utterance_id": "LA_E_9332881", "speaker_id": "LA_0009", "codec": "alaw", "transmission": "ita_tx", "attack_id": "A07", "trim": "notrim", "phase": "eval"}

Quick Start

python

from datasets import load_dataset

ds = load_dataset("SpeechAntiSpoofingBenchmarks/ASVspoof2021_LA", split="test")
print(ds[0])

Stats

Stat	Value
Total trials	181,566
Bonafide	18,452
Spoof	163,114

Phase breakdown (in notes.phase): eval 148,176 · progress 16,464 · hidden 16,926.

Source provenance

—Original dataset: https://www.asvspoof.org/index2021.html
—Evaluation protocol: protocols/ASVspoof2021.LA.cm.eval.trl.txt

Evaluation

For evaluation instructions and submission format, see `submissions/README.md`.

Citation

Original paper: https://arxiv.org/abs/2109.00537

bibtex

@inproceedings{yamagishi21_asvspoof,
  title     = {{ASVspoof 2021: accelerating progress in spoofed and deepfake speech detection}},
  author    = {Junichi Yamagishi and Xin Wang and Massimiliano Todisco and Md Sahidullah and Jose Patino and Andreas Nautsch and Xuechen Liu and Kong Aik Lee and Tomi Kinnunen and Nicholas Evans and Héctor Delgado},
  year      = {2021},
  booktitle = {{2021 Edition of the Automatic Speaker Verification and Spoofing Countermeasures Challenge}},
  pages     = {47--54},
  doi       = {10.21437/ASVSPOOF.2021-8},
}

Maintainer

Maintained by Kirill Borodin (SpeechAntiSpoofingBenchmarks).

—Email: ~~k.n.borodin@mtuci.ru~~ (deprecated — use kborodin.research@gmail.com)
—Telegram: @korallll_ai