The database GTZAN+

Short Description

GTZAN+ is a music database composed of 15 music genres. It is an extension of the GTZAN database composed of 10 music genres (available here) augmented by the 5 following Afro music genres: Bikutsi, Makossa, Bamileké, Salsa and Zouk. Each genre in GTZAN+ is represented by 100 ".wav" files of 30 seconds each (22050 Hz, 32-bit floating point, mono). Bikutsi, Makossa and Bamileke are Cameroonian traditional music genres. Zouk is French-Carabbean music genre and Salsa is a music genre having Cuban origins, but many salsa songs played by Cameroonian singers have been included in the database.

Download GTZAN+

GTZAN+ (15 genres)
GTZAN (10 genres) AFRO (5 genres)
blues classical country disco hiphop jazz metal pop reggae rock bikutsi makossa bamileke salsa zouk

Related Papers

Hierarchical classification experiments have yet been realized on GTZAN+ in the 3 papers listed below. Paper [1] obtained an accuracy of 85.2%, paper [2] exhibited an accuracy of 92%, and paper [3] showed an accuracy of 97%.

  1. S. Iloga, O. Romain and M. Tchuenté, "A sequential pattern mining approach to design taxonomies for hierarchical music genre recognition" , Pattern Analysis and Applications, DOI 10.1007/s10044-016-0582-7, Vol. 21(2), pp. 363-380, Springer, 2016. (Download this paper)

  2. S. Iloga, O. Romain and M. Tchuenté, "An accurate HMM-based similarity measure between finite sets of histograms" , Pattern Analysis and Applications, DOI 10.1007/s10044-018-0734-z, Vol. 22(3), pp. 1079-1104, Springer, 2018. (Download this paper)

  3. S. Iloga, O. Romain and M. Tchuenté, "An efficient generic approach for automatic taxonomy generation using HMMs" , Pattern Analysis and Applications, DOI 10.1007/s10044-020-00918-0, pp. 1-22, Springer, 2020. (Download this paper)

  4. Coming soon ...

Download ARFF files

In order to perform music genre classification on the content of GTZAN+ in public tools like WEKA or MEKA, an ARFF (Attribute-Relation File Format) file is required. This file contains the music descriptors used to describe each song of the database. A detailed description of the ARFF file format is available here. Three ARFF files used in the related papers [1,2,3] can be downloaded from this page, 2 files for GTZAN and 1 file for GTZAN+. In each file, the songs of the corresponding database are individually described by 34 timbre and 401 rhythm music descriptors. These 3 files are available through the following links:

  • ARFF file of GTZAN used in WEKA: download

  • ARFF file of GTZAN+ used in WEKA: download

  • ARFF file of GTZAN corresponding to the taxonomy T4 in the related paper [1] used in MEKA: download. This taxonomy is presented in Figure 1.

    Taxonomy

    Figure 1: The taxonomy T4 of GTZAN+ in the related paper [1]

Author contact

  • Name: Pierre Sylvain Iloga Biyik
  • Email: sylvain.iloga@gmail.com , sylvain.iloga@ensea.fr , pierre.iloga-biyik@cyu.fr
  • Professional adress: Higher Teacher's Training College, Department of Computer Science,University of Maroua, PO.Box 55, Maroua,Cameroon.