Papers of the day   All papers

FSD50K: an Open Dataset ofHuman-Labeled Sound Events

Comments

Eduardo Fonseca: 🔊Happy to announce FSD50K: the new open dataset of human-labeled sound events! Over 51k Freesound audio clips, totalling over 100h of audio manually labeled using 200 classes drawn from the AudioSet Ontology. Paper: https://arxiv.org/pdf/2010.00475.pdf Dataset: http://doi.org/10.5281/zenodo.4060432 https://t.co/oKzW55LGWp

7 replies, 267 likes


Justin Salamon: Exciting new dataset for audio ML research

2 replies, 32 likes


Daniel Situnayake #BlackLivesMatter: This looks like an excellent dataset for common environmental sounds!

0 replies, 15 likes


arxiv: FSD50K: an Open Dataset of Human-Labeled Sound Events. http://arxiv.org/abs/2010.00475 https://t.co/n2fEdG1fGt

0 replies, 15 likes


Alexandre Défossez: 100h of all kind of sounds, human labeled and with clear licensing. That's exciting! 🤖

1 replies, 14 likes


AV Speech Processing: FSD50K, an open dataset (108.3 hours) of human-labeled sound events- nice! See https://arxiv.org/abs/2010.00475 Content has 200 classes - see https://annotator.freesound.org/fsd/release/FSD50K/ Framework for sound event classification will be here https://github.com/edufonseca/FSD50K_baseline when @edfonseca_ cleans the code 😉

0 replies, 11 likes


AK: FSD50K: an Open Dataset of Human-Labeled Sound Events pdf: https://arxiv.org/pdf/2010.00475.pdf abs: https://arxiv.org/abs/2010.00475 https://t.co/6lgVS13DBo

0 replies, 11 likes


Sajjad Abdoli (@🏡): Amazing dataset of sound events with nice annotations. Something that I was waiting for!

1 replies, 9 likes


Jonathan Le Roux: This is exciting 🎉

1 replies, 3 likes


Underfox: In this paper is presented FSD50K, an open dataset containing over 51k audio clips totalling over 100h of audio manually labeled using 200 classes drawn from the AudioSet Ontology. #MachineLearning https://arxiv.org/pdf/2010.00475.pdf https://t.co/Tu3UW1qZpe

0 replies, 2 likes


Edward Dixon: The original AudioSet was larger, but based on YouTube videos, really painful to download. Friends working on language identification models have been struggling with background noise, this should be an excellent data augmentation resource.

0 replies, 2 likes


L Quera: 🎉New dataset! 🎉

0 replies, 2 likes


Khaled K: Super interesting!

0 replies, 2 likes


Emilia Gomez: A great contribution from @edfonseca_ et al. @freesounddev team @mtg_upf to the research community #audio #tags #sound #opensience Not to miss!

0 replies, 2 likes


Joseph O'Brien Antognini: I have been looking forward to this for a *very* long time! (As has everyone else on ML at Whisper!) This is a tremendous contribution to the audio classification field!

0 replies, 1 likes


Jordi Pons: The best of FSD50K is that you can easily navigate through it! You can listen to the audio, report errors and check its annotations online: 🔗 https://annotator.freesound.org/fsd/release/FSD50K/

1 replies, 0 likes


Content

Found on Oct 02 2020 at https://arxiv.org/pdf/2010.00475.pdf

PDF content of a computer science paper: FSD50K: an Open Dataset ofHuman-Labeled Sound Events