FMA90k: A Multi-Label Dataset Extracted from the Free Music Archive
Rodolfo M. Pereira¹ , Yandre M. G. Costa² and Carlos N. Silla Jr.¹
¹Computer Music Technology Laboratory – Graduate Program in Computer Science
Pontifical Catholic University of Paraná (PUCPR), Curitiba, PR, Brazil
²Graduate Program in Computer Science – Informatics Department
State University of Maringá (UEM), Maringá, PR, Brazil
Email: {rodolfomp123, yandre.costa, carlos.sillajr}@gmail.com
Full dataset with full lenght songs:
JSON files used to find the songs metadata:
ARFF files used in the FMA90k experiments:
Feature files extracted from the FMA90k songs:
To allow other researchers comparison, we provide text files with the songs IDs divided into the same Folds used in the experiments:
Scripts used to Download FMA90k and its metadata, such as used to generate the songs features (with YAAFE framework):
Single-Label subset with full lenght songs:
JSON files used to find the FMA-SL songs metadata:
ARFF files used in the FMA-SL experiments:
Feature files extracted from the FMA-SL songs: