The MTG-UPF, through Freesound, and the Google Machine Perception Team, 
creators of  AudioSet Ontology, joint efforts to create an open audio dataset 
and to organize a machine learning competition, efforts aimed to stimulate 
research in sound event recognition. The MTG-UPF has been awarded a Google 
Faculty Research Award to support this initiative.

The Freesound Datasets platform, developed by the MTG, is an open platform for 
the creation of audio collections labeled by people and based on Freesound. 
With the collaboration of the Google Machine Perception Team, we have created 
the first dataset, FSD, which is annotated using the AudioSet Ontology. 

In order to promote the use of FSD, we have created the Freesound 
General-Purpose Audio Tagging Challenge. In this challenge the participants are 
asked to build systems able to recognize over 40 different categories of 
everyday sounds (a small subset of FSD). This competition is hosted in Kaggle 
(a well-known platform for machine learning competitions) under the framework 
of the DCASE Challenge (an academic competition featuring tasks related to the 
computational analysis of sound events).

More info:

Xavier Serra
Music Technology Group
Universitat Pompeu Fabra, Barcelona

dupswapdrop: music-dsp mailing list

Reply via email to