Record Details

Meter2800

Harvard Dataverse (Africa Rice Center, Bioversity International, CCAFS, CIAT, IFPRI, IRRI and WorldFish)

View Archive Info
 
 
Field Value
 
Title Meter2800
 
Identifier https://doi.org/10.7910/DVN/0CLXBQ
 
Creator Abimbola, Jeremiah
Daniel Kostrzewa
Pawel Kasprowski
 
Publisher Harvard Dataverse
 
Description Context
Music enthusiasts have been trying for a long time to understand how to determine the time signature of any music track and what differentiates one song from another in terms of the rhythmic characteristics.

This data hopefully can give the opportunity to do just that.



Content:

A collection of zipped files. FMA.tar.gz for FMA-medium dataset, MAG.tar.gz for MagnaTagATune dataset, OWN.tar.gz for audio files curated by the authors. The GTZAN dataset was also used in this annotation process but has been excluded in the content because it can be downloaded from Kaggle.



4 CSV files - Containing train and test data annotations of the audio files performed by a professional musician. The audio files were annotated into 4 meter classes (3, 4, 5 and 7). Where 3, 4 are regular meters and 5, 7 are irregular meters. Meters 3 and 4 used 1200 audio files each while 5 and 7 used 200 audio files each, totalling 2800 audio files. Each audio file is 30 seconds long.


Description:

The train /validation /test split is done in the ration 60:15:25 where all the classes are equally represented in each data file. Hence the file data_train_4_classes contains 1,680 records for all 4 meter classes, the file data_val_4_classes contains 420 records for all 4 meter classes while the file data_test_4_classes contains 700 records for all 4 meter classes. The file data_train_2_classes contains 1,440 records for only two meter classes (3 and 4) , the file data_val_2_classes contains 360 records for only two meter classes and the file data_test_2_classes contains 600 records for only two meter classes.
 
Subject Arts and Humanities
Computer and Information Science
Music Time Signature
 
Date 2022-12-14
 
Contributor Abimbola, Jeremiah