What question did this study set out to answer?

To overcome the scarcity of large, open-access audio collections for feature learning and end-to-end learning by introducing a standardized music information retrieval dataset.

December 6, 2016Open Access

FMA: A Dataset For Music Analysis

View Full Paper

Ask AI

Bookmark

Authors

MDMichaël DefferrardLaboratoire des signaux et systèmes KBKirell BenziÉcole Polytechnique Fédérale de Lausanne PVPierre VandergheynstHaute École Pédagogique du Canton de Vaud

Discussion

Member takes

Implication

Benchmark study reveals an open 106,574-track music dataset with hierarchical genre taxonomies, highlighting reproducible evaluation standards for music information retrieval.

Key Points

To overcome the scarcity of large, open-access audio collections for feature learning and end-to-end learning by introducing a standardized music information retrieval dataset.
Curated 917 GiB (343 days) of Creative Commons-licensed audio containing 106,574 tracks from 16,341 artists and 14,854 albums.
Organized audio tracks into a hierarchical taxonomy of 161 genres alongside pre-computed features, track- and user-level metadata, tags, and text biographies.
Defined standard train/validation/test splits across three sub-collections and established baseline models for genre recognition tasks.
Established an open-access multi-genre audio repository with 106,574 full-length tracks, standardized subsets, and high-dimensional acoustic feature sets.
Demonstrated baseline genre classification performance to validate the benchmark's utility for comparative music information retrieval tasks.

Cite This Study

Defferrard et al. (2016) studied this question.

synapsesocial.com/papers/6a11e847c031bb6829a58bab https://doi.org/10.48550/arxiv.1612.01840

View Full Paper

Ask AI

Bookmark

Implication

Benchmark study reveals an open 106,574-track music dataset with hierarchical genre taxonomies, highlighting reproducible evaluation standards for music information retrieval.

FMA: A Dataset For Music Analysis

Discussion

Member takes

Implication

Key Points

Cite This Study

FMA: A Dataset For Music Analysis

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider