HEAR NeurIPS 2021 Datasets (Holistic Evaluation of Audio Representations)

NOTE: The datasets have different open licenses. Please see LICENSE.txt for each individual dataset's license.

These are the evaluation tasks for the HEAR (Holistic Evaluation of Audio Representations) 2021 NeurIPS challenge.

The aim of this challenge is to develop a general-purpose audio representation that provides a strong basis for learning in a wide variety of tasks and scenarios. The HEAR 2021 challenge invites you to create an audio embedding that is as holistic as the human ear, i.e., one that performs well across a variety of everyday domains: What approach best generalizes to a wide range of downstream audio tasks without fine-tuning?

HEAR 2021 evaluates audio representations using a benchmark suite across a variety of domains, including speech, environmental sound, and music.

For more information, see the HEAR 2021 website and upcoming PMLR journal article.

Datasets were all normalized to a common human-readable format using hearpreprocess. This Zenodo archive has all audio resampled to 48000Hz. For other sampling rates (16000, 22050, 32000, 44100), please download files (requester pays) from Google Storage gs://hear2021-archive/tasks/ or AWS s3://hear2021-archive/tasks/

The TFDS datasets (crema-d, gztan genre, and gztan music/speech) had a preprocessing bug and are retracted. We will release fixed versions of these datasets.