MIRTracks: A Large-Scale Multi-Dimensional Multi-Track Music Dataset

Abstract

This paper presents MIRTracks, a large-scale dataset containing 240 h of royalty-free multi-track audio, aiming to address the limitations of traditional music source separation datasets, including single-dimensional annotation and semantic information gaps. By integrating multi-dimensional musical information annotation with a semi-automated annotation pipeline, MIRTracks achieves high- quality semantic annotation across rock, electronic, and pop music genres. Experiments demonstrate that fine-tuning a small-scale model on this dataset significantly improves beat detection accuracy from 66.2% to 80.1%, reaching 91.0% of the performance of large-scale models

Keywords

dataset / annotation / music information retrieval

Author Contributions

Y.Q.: conceptualization, supervision; Y.L.: software, data curation, writing—original draft preparation, methodology. All authors have read and agreed to the published version of the manuscript.

Funding

This research was supported by the project entitled “Multimodal Emotion Analysis and Mapping Based on Music-Visual Synesthesia”.

Data Availability Statement

All methods for obtaining research data and annotation tools have been fully open-sourced at: https://github.com/bigblackLee123/MIRTracks.git.

Conflicts of Interest

The authors declare no conflict of interest.

Use of AI and AI-Assisted Technologies

During the preparation of this work, the authors used Cursor to assist with code writing, and used Doubao to assist with translation work. After using these tools/services, the authors reviewed and edited the content as needed and takes full responsibility for the content of the published article.

References

[1] Rafii Z.; Liutkus A.; Stöter F.R.; et al. MUSDB18—A corpus for music separation. arXiv preprint 2017, arXiv:1710.11192.

[2] Stöter F.R.; Uhlich S.; Liutkus A.; et al. Open-Unmix: A Reference Implementation for Music Source Separation. J. Open Source Softw. 2019, 4, 1667.

[3] Tong W.; Zhu J.; Chen J.; et al. SCNet: Sparse compression network for music source separation. arXiv preprint 2024, arXiv:2401.13276.

[4] Bittner R.M.; Salamon J.; Tierney M.; et al. A Multitrack Dataset for Annotation-Intensive MIR Research. In Proceedings of the 15th International Society for Music Information Retrieval Conference (ISMIR), Taipei, Taiwan, 27- 31 October 2014; pp. 155-160.

[5] Hadjeres G.; Pachet F.; Nielsen F. DeepBach: A Steerable Model for Bach Chorales Generation. In Proceedings of the 34th International Conference on Machine Learning (PMLR), Sydney, Australia, 6-11 August 2017; pp. 1362-1371.

[6] Dhariwal P.; Jun H.; Payne C.; et al. Jukebox: A generative model for music. arXiv e-print 2020, arXiv:2005.00341.

[7] Agostinelli A.; Denk T.I.; Borsos Z.; et al. MusicLM: Generating music from text. arXiv preprint 2023, arXiv:2301.11325.

[8] Foscarin F.; Schlüter J.; Widmer G. Beat this! Accurate beat tracking without DBN postprocessing. arXiv preprint 2024, arXiv:2407.21658.

[9] Sturm B.L. The GTZAN dataset: Its contents, its faults, their effects on evaluation, and its future use. arXiv preprint 2013, arXiv:1306.1461.

[10] Liutkus, A.; Stöter F.-R.; Rafii Z.; et al. The 2016 Signal Separation Evaluation Campaign. In Latent Variable Analysis and Signal Separation, Proceedings of the 12th International Conference, LVA/ICA 2015, Liberec, Czech Republic, 25- 28 August 2015; pp. 323-332;Tichavský P., Babaie-Zadeh, M., Michel, O.J.J.; et al., Eds.; Springer International Publishing: Cham, Switzerland, 2017. https://doi.org/10.1007/978-3-319-19544-2_31.