Collection  |  Manuscript/Mixed Material FMA : a dataset for music analysis. Free music archive

[ Small version of FMA data ]

More Resources

[ FMA git repo ]
[ FMA metadata ]

About this Item

Title
FMA : a dataset for music analysis.
Other Title
Free music archive
Contributor Names
WFMU (Radio station : Montclair, N.J.)
Created / Published
2017-
Subject Headings
-  Musical analysis
Genre
Sound recordings
Notes
-  The full dataset includes over 100,000 ful length MP3 audio recordings, and truncated files. There are four .csv metadata files that include: track-level information (such as ID, title, artist, genres, tags and play counts), genre IDs, data generated using LibRosa (a package for music and audio analysis), and data generated by echonest (applicable to a subset of the records). A published paper is a single PDF document, and code (notebooks and scripts) that have been developed for interacting with the dataset and metadata files. The Library is currently only able to provide dowload access for the smallest version of the dataset. Visit the GitHub archive in order to download the larger versions.
-  Description based on GitHub page, Sept. 25, 2018.
Medium
dataset
Repository
s-Online Electronic Resource
Digital Id
https://github.com/mdeff/fma External
https://hdl.loc.gov/loc.gdc/gdcdatasets.2018655052_small
https://hdl.loc.gov/loc.gdc/gdcdatasets.2018655052_gdcfma-gitrepo
https://hdl.loc.gov/loc.gdc/gdcdatasets.2018655052_gdcfma-metadata
Library of Congress Control Number
2018655052
Rights Advisory
The majority of the audio in the FMA is licensed as either Creative Commons or Public Domain, while a smaller portion are under less common licenses. The code is available under the MIT license, and the paper and metadata files are released under the CC BY 4.0 license.
Language
English
Online Format
compressed data
Description
The full dataset includes over 100,000 ful length MP3 audio recordings, and truncated files. There are four .csv metadata files that include: track-level information (such as ID, title, artist, genres, tags and play counts), genre IDs, data generated using LibRosa (a package for music and audio analysis), and data generated by echonest (applicable to a subset of the records). A published paper is a single PDF document, and code (notebooks and scripts) that have been developed for interacting with the dataset and metadata files. The Library is currently only able to provide dowload access for the smallest version of the dataset. Visit the GitHub archive in order to download the larger versions. Description based on GitHub page, Sept. 25, 2018.
LCCN Permalink
https://lccn.loc.gov/2018655052
Additional Metadata Formats
MARCXML Record
MODS Record
Dublin Core Record

Rights & Access

The Library of Congress is providing access to The Selected Datasets Collection for educational and research purposes. The Library has obtained permission for the use of many materials in the Collection, and presents additional materials for educational and research purposes in accordance with fair use under United States copyright law. Researchers should watch for modern documents that may be copyrighted (for example, published in the United States more than 95 years ago, or unpublished and the author died less than 70 years ago).

You are responsible for deciding whether your use of the items in this collection is legal. You are also responsible for securing any permissions needed to use the items. You will need written permission from the copyright owners of materials not in the public domain for distribution, reproduction, or other use of protected items beyond that allowed by fair use or other statutory exemptions. Some content may be protected under international law. You may also need permission from holders of other rights, such as publicity and/or privacy rights.

More about Copyright and other Restrictions

Credit Line: Library of Congress, Digital Collections Management and Services Division

Cite This Item

Citations are generated automatically from bibliographic data as a convenience, and may not be complete or accurate.

Chicago citation style:

Wfmu. FMA: a dataset for music analysis. 2017. Manuscript/Mixed Material. https://www.loc.gov/item/2018655052/.

APA citation style:

Wfmu. (2017) FMA: a dataset for music analysis. [Manuscript/Mixed Material] Retrieved from the Library of Congress, https://www.loc.gov/item/2018655052/.

MLA citation style:

Wfmu. FMA: a dataset for music analysis. 2017. Manuscript/Mixed Material. Retrieved from the Library of Congress, <www.loc.gov/item/2018655052/>.

More Collections like this