Key facts
Contents & fields
The original dataset contains at least 1700 audio clips (mp3 format) of different music genres, each about 270-300 seconds long. The database is divided into 17 genres, each with an annotation file containing genre classification labels. The processed dataset consists of spectrogram slices converted from all audio, with three levels of labels: Level 1 (2-class: classical vs. non-classical); Level 2 (9-class: Symphony, Opera, Solo, Chamber, Pop, Dance_and_house, Indie, Soul_or_r_and_b, Rock); Level 3 (16-class: partially listed).
- Level 1 label — 2-class: classical vs. non-classical
- Level 2 label — 9-class: 3_Symphony, 4_Opera, 5_Solo, 6_Chamber, 7_Pop, 8_Dance_and_house, 9_Indie, 10_Soul_or_r_and_b, 11_Rock
- Level 3 label — 16-class: 3_Symphony, 4_Opera, 5_Solo, 6_Chamber, ... (others not listed)
Research uses
Suitable for music genre classification, audio signal processing, and machine learning classification model training.
This card was drafted from the source page; institution, coverage, time span, scale, fields and license are subject to the official page (pending human review).
Keywords
Why this is hard to get on your own
In music genre classification tasks, there is a lack of publicly available datasets with clear and hierarchical annotations. This dataset provides a three-level label system for multi-granularity classification research.