Dataset Overview

Basic Information

• Name: Music Genre Dataset
• License: MIT
• Languages: Chinese, English
• Tags: music, art
• Size: 10K<n<100K

Description

• Overview: Contains about 1,700 music pieces in .mp3 format, each 270–300 seconds long, sampled at 22 kHz. The pieces are divided into 16 distinct music styles.
• Source: Data sourced from NetEase Music; genre tags are included with the downloads.
• Classification: 16 music styles.

Structure

• Audio Format: .mp3
• Sample Rate: 22 kHz
• Duration Range: 270–300 seconds
• Label Taxonomy: Three‑level hierarchy (2 coarse classes, 9 middle classes, 16 fine‑grained classes).

Usage Example

• Loading: Use the load_dataset function; both eval and default subsets are available.
• Processing: The dataset includes train, validation, and test splits and supports audio‑classification tasks.

Creation

• Collection & Annotation: Collected and annotated by CCMUSIC students; 1,700 pieces grouped into 17 styles.
• Copyright: Only spectrograms are provided due to copyright restrictions.

Notes

• Language Bias: Majority of tracks are English.
• Sample Balance: Class distribution is imbalanced.

License

• Type: MIT License
• Copyright Holder: CCMUSIC
• Conditions: Free use, copy, modify, merge, publish, distribute, sublicense, and sell copies, provided the copyright notice and license terms are retained.

Citation

• Authors: Monan Zhou, Shenyang Xu, Zhaorui Liu, Zhaowen Wang, Feng Yu, Wei Li, Baoqiang Han
• Title: CCMusic: an Open and Diverse Database for Chinese and General Music Information Retrieval Research
• Year: 2024
• Version: 1.2
• URL: https://huggingface.co/ccmusic-database