Metrical structure annotations for the GTZAN dataset

We present a dataset consisting of metrical structure annotations for the GTZAN dataset.

  • The metrical structure is represented by the pulse rates (in BPM) of all the metrical levels present in the music
  • The annotations have been made by Professional drummers
  • Each track of the dataset have been annotated at least twice, so that the level of inter-annotator agreement can be assessed

Download it here

Content of the dataset:

  • READ_ME.txt: contains useful information about the dataset, such as a description of the data structure, citation information etc.
  • annotations_MetricalStructure_GTZAN.json: contains the annotations for all the tracks
  • csv: contains the annotations in csv format. One file per track, containing all the annotations.

If you make use of this dataset, please cite:
[1] E. Quinton, C. Harte, and M. Sandler, “Extraction of Metrical Structure from Music Recordings,” in Proc. of the 18th Int. Conference on Digital Audio Effects (DAFx). Trondheim, Norway, Nov 30 - Dec 3, 2015.
Read paper here

The corresponding audio files can be found here