I have collected several music datasets, especially for Chinese music:

• MGD: Large Scale Chinese Folk Songs Dataset (2014-2023)

  

  Over 31000 Chinese folk songs in symbolic format.

• MGDplus (2021-2022)

  

  1214 Selected symbolic Han Chinese folk songs from seven regions with five annotations, partially audio-score aligned. Some examples can be found here.

• Music Emotion Dataset (2021-2022)

  Music Emotion Recognition (MER) has recently received considerable attention. We developed the MER1101 Music Emotion Dataset to support the MER research requiring large music content libraries. It is a dataset that explores the relationship between music and human emotion. It is based on Russell's valence-arousal emotion model and contains 1101 music audio gathered from the internet, each ranging from 16.5 seconds to 125.5 seconds.

• Singing Quality Evaluation Dataset (2023-2024)

  We have developed a new dataset for Singing Quality Evaluation, comprising over 2,000 Chinese popular songs. Each song in the dataset is rated across eight sub-dimensions, along with an overall score, providing a comprehensive assessment of singing quality.

• End-to-end Comprehensive Music Quality Evaluation (2024-2025, ongoing)
  To be announced