English

MAP-Music2Vec: A Simple and Effective Baseline for Self-Supervised Music Audio Representation Learning

Sound 2022-12-07 v1 Artificial Intelligence Machine Learning Multimedia Audio and Speech Processing

Abstract

The deep learning community has witnessed an exponentially growing interest in self-supervised learning (SSL). However, it still remains unexplored how to build a framework for learning useful representations of raw music waveforms in a self-supervised manner. In this work, we design Music2Vec, a framework exploring different SSL algorithmic components and tricks for music audio recordings. Our model achieves comparable results to the state-of-the-art (SOTA) music SSL model Jukebox, despite being significantly smaller with less than 2% of parameters of the latter. The model will be released on Huggingface(Please refer to: https://huggingface.co/m-a-p/music2vec-v1)

Keywords

Cite

@article{arxiv.2212.02508,
  title  = {MAP-Music2Vec: A Simple and Effective Baseline for Self-Supervised Music Audio Representation Learning},
  author = {Yizhi Li and Ruibin Yuan and Ge Zhang and Yinghao Ma and Chenghua Lin and Xingran Chen and Anton Ragni and Hanzhi Yin and Zhijie Hu and Haoyu He and Emmanouil Benetos and Norbert Gyenge and Ruibo Liu and Jie Fu},
  journal= {arXiv preprint arXiv:2212.02508},
  year   = {2022}
}
R2 v1 2026-06-28T07:22:48.216Z