LAION-Audio-630K Freesound Dataset

LAION-Audio-630K is the largest audio-text dataset publicly available and a magnitude larger than previous audio-text datasets (by 2022-11-05). Notably, it combines eight distinct datasets, which includes the Freesound dataset.

Specifically, this Hugging face repository contains two versions of Freesound dataset. Details of each dataset (e.g. how captions are made etc.) could be found in the "datacard" column of the table below.

Freesound (full): The complete Freesound dataset, available at /freesound folder.
Freesound (no overlap): Made based on Freesound(full), with samples from ESC50, FSD50K, Urbansound8K and Clotho removed. available at /freesound_no_overlap folder.

As of the structure and format of freesound and freesound_no_overlap folder, please refer to this page.

Name	Duration	Number of Samples	Data Type	Metadata	Data Card
Freesound (no overlap)	2817.31hrs	460801	1-2 captions per audio, audio	website csv	data card
Freesound (full)	3033.38hrs	515581	1-2 captions per audio, audio	website csv	data card

Metadata csv file

For each of the two datasets, we provide a metadata csv file including the following columns:

audio_filename: The filename of the audio file in .tar files. exemple: 2394.flac
caption_i: the i-th caption of the audio file
freesound_id: The freesound id of the audio file.
username: The username of the uploader of the audio file.
freesound_url: The url of the audio file in freesound.org
username: The freesound username of the uploader of the audio file.
license: The license of the audio file. http://creativecommons.org/licenses/by/3.0/

Credits & Licence

!!!TERM OF USE!!!: By downloading files in this repository, you agree that you will use them for research purposes only . If you want to use Freesound clips in LAION-Audio-630K for commercial purposes, please contact Frederic Font Corbera at [email protected].

Freesound Credit:

All audio clips from Freesound are released under Creative Commons (CC) licenses, while each clip has its own license as defined by the clip uploader in Freesound, some of them requiring attribution to their original authors and some forbidding further commercial reuse. Specifically, here is the statistics about licenses of audio clips involved in LAION-Audio-630K:

License	Number of Samples
http://creativecommons.org/publicdomain/zero/1.0/	260134
https://creativecommons.org/licenses/by/4.0/	97090
http://creativecommons.org/licenses/by/3.0/	89337
http://creativecommons.org/licenses/by-nc/3.0/	31680
https://creativecommons.org/licenses/by-nc/4.0/	26736
http://creativecommons.org/licenses/sampling+/1.0/	11116

Acknowledgement

The whole collection process as well as all usage of the LAION-Audio-630K are conducted by Germany non-profit pure research organization LAION. All contributors and collectors of the dataset are considered as open source contributors affiliated to LAION. These community contributors (Discord ids) include but not limited to: @marianna13#7139, @Chr0my#0173, @PiEquals4#1909, @Yuchen Hui#8574, @Antoniooooo#4758, @IYWO#9072, krishna#1648, @dicknascarsixtynine#3885, and @turian#1607. We would like to appreciate all of them for their efforts on the LAION-Audio-630k dataset.