Published December 17, 2022 | Version v2
Dataset Open

FSD-FS

  • 1. Queen Mary University of London
  • 2. Amazon Alexa

Description

FSD-FS is a publicly-available database of human labelled sound events for few-shot learning. It spans across 143 classes obtained from the AudioSet Ontology and contains 43805 raw audio files collected from the FSD50K. FSD-FS is curated at the Centre for Digital Music, Queen Mary University of London.

Citation

If you use the FSD-FS dataset, please cite our paper and FSD50K.

@article{liang2022learning,
  title={Learning from Taxonomy: Multi-label Few-Shot Classification for Everyday Sound Recognition},
  author={Liang, Jinhua and Phan, Huy and Benetos, Emmanouil},
  journal={arXiv preprint arXiv:2212.08952},
  year={2022}
}

@ARTICLE{9645159,  author={Fonseca, Eduardo and Favory, Xavier and Pons, Jordi and Font, Frederic and Serra, Xavier},  journal={IEEE/ACM Transactions on Audio, Speech, and Language Processing},   title={FSD50K: An Open Dataset of Human-Labeled Sound Events},   year={2022},  volume={30},  number={},  pages={829-852},  doi={10.1109/TASLP.2021.3133208}}

About FSD-FS

FSD-FS is an open database for multi-label few-shot audio classification containing 143 classes drawn from the FSD50K. It also inherits the AudioSet Ontology. FSD-FS follows the ratio 7:2:1 to split classes into base, validation, and evaluation sets, so there are 98 classes in the base set, 30 classes in the validation set, and 15 classes in the evaluation set (More details can be found in our paper).

LICENSE

FSD-FS are released in Creative Commons (CC) licenses. Same as FSD50K, 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. For more details, ones can refer to the link.

FILES

FSD-FS are organised in the structure:

root
|
└─── dev_base
|
└─── dev_val
|
└─── eval

 

REFERENCES AND LINKS

[1] Gemmeke, Jort F., et al. "Audio set: An ontology and human-labeled dataset for audio events." 2017 IEEE international conference on acoustics, speech and signal processing (ICASSP). IEEE, 2017. [paper] [link]

[2] Fonseca, Eduardo, et al. "Fsd50k: an open dataset of human-labeled sound events." IEEE/ACM Transactions on Audio, Speech, and Language Processing 30 (2021): 829-852. [paper] [code]

Files

dev_base.zip

Files (27.6 GB)

Name Size Download all
md5:fa4b0ac3d7cbdb282f43c9e70b98ce03
1.1 GB Download
md5:2e7d227154d7fa84719b6fafd6c9e86b
1.1 GB Download
md5:ee92ef7019a31834600daa29c7731514
1.1 GB Download
md5:e412b3b053cec2ea5e27307e5a5c776e
1.1 GB Download
md5:b1eb4a54dd8d3f454248c8e22c3f7883
1.1 GB Download
md5:1e55fce8260251ed0cae08ef7576a93d
1.1 GB Download
md5:a46f723b229a79d45fbb46646dfcb80a
1.1 GB Download
md5:b3a44fd8d8489dc54e3b45dec5976c62
1.1 GB Download
md5:3aa5730b30a44ce97f4f4299876b0f26
1.1 GB Download
md5:0499a000205c6e2213c6e419f5e0523f
1.1 GB Download
md5:001f841dd0a64067a21bc89a138cfb19
1.1 GB Download
md5:12ed26eed67fcb984c123c49397e442f
821.3 MB Preview Download
md5:9fd5551cf4f4fb587cd81e837d64d045
1.1 GB Download
md5:d5a3928ed74eb4938fdc378b555373d8
1.1 GB Download
md5:eba5c9ad1ffcd6db6fa20a0f5901f84c
1.1 GB Download
md5:70cccc8a6b38dc573bd250e1a4b63236
1.1 GB Download
md5:2818ddb55b174081eaaf815de3a1a313
1.1 GB Download
md5:78ae71a178cee4d362c8741e929c2799
107.4 MB Preview Download
md5:aa5c96dfe024d270435b27e2f9961f70
1.1 GB Download
md5:13ad455a4013f113500ff0d7a94b8ce7
1.1 GB Download
md5:8df5c1c5e619eaf3f6ff9514ac85df21
885.7 MB Preview Download
md5:e307f28717c0ecb3c6d2bd843603b744
314.0 kB Preview Download
md5:cf845f631bccfa135bf571e550d350e4
2.1 GB Download
md5:3ea7f758ad3bc3611a3ff5987546d604
2.1 GB Download
md5:f744ff90202081014b7131a393133af2
2.1 GB Preview Download

Additional details

Funding

DTP 2020-2021 Queen Mary University of London EP/T518086/1
UK Research and Innovation