cheyenne_h (FMA Admin)
cheyenne_h on 05/22/2017 at 06:15AM
A Music Information Retrieval Dataset, Made With FMA
You may recall some news we shared last summer about a music dataset that was in progress - now it's complete!
Michaël Defferrard, Kirell Benzi, Pierre Vandergheynst & Xavier Bresson, a team of researchers interested in MIR (music information retrieval), have put together some audio datasets using the FMA's collection! They have written a paper about the FMA's potential as a viable and legal option in digital audio research. Since recorded music is often fraught with copyright concerns, machine learning sets for music have been hard to find, or not sufficient for some projects. You can find the datasets on GitHub.
The FMA's dedication to the Creative Commons philosophy of sharing has been a boon for artists, educators, and researchers - as well as, of course, listeners! Hopefully this dataset will be widely adopted for use by MIR researchers. It seems to have already been well-received via Twitter:
A great big THANKS go to this team for thinking of FMA and sharing our collection with a new audience, for a new purpose, in a new way! We look forward to the cool projects that will stem from their efforts.
![Share](https://web.archive.org/web/20190907182116im_/https://s7.addthis.com/static/btn/sm-share-en.gif)