---
tags:
- espnet
- audio
- self-supervised-learning
- beats
datasets:
- audioset
license: cc-by-4.0
arxiv: 2212.09058
---

# BEATs base encoder — AudioSet-2M, iteration 0

BEATs base audio encoder pre-trained on AudioSet-2M with ESPnet, reproducing
[BEATs](https://arxiv.org/abs/2212.09058) (Table 4). Iteration 0 is trained against
random-projection targets. Produced by the [egs2/audioset/ssl1](https://github.com/espnet/espnet/tree/master/egs2/audioset/ssl1) recipe.

## Usage

Load into an ESPnet downstream classification recipe (e.g. `egs2/esc50/asr1`,
`egs2/as20k/cls1`) via `beats_ckpt_path`:

```yaml
encoder: beats
encoder_conf:
    beats_ckpt_path: beats_encoder_iter0.pt
    fbank_mean: 15.66439
    fbank_std: 6.38312
```

## Downstream results

| | AS-20K test mAP | ESC-50 (5-fold acc) |
|---|---|---|
| iter 0 | 29.72 | 93.35 |

## Citation

```bibtex
@inproceedings{chen2022beats,
  title={BEATs: Audio Pre-Training with Acoustic Tokenizers},
  author={Chen, Sanyuan and Wu, Yu and Wang, Chengyi and Liu, Shujie and
          Tompkins, Daniel and Chen, Zhuo and Wei, Furu},
  booktitle={ICML}, year={2023}
}
@inproceedings{bharadwaj2025openbeats,
  title={OpenBEATs: A Fully Open-Source General-Purpose Audio Encoder},
  author={Bharadwaj, Shikhar and Cornell, Samuele and Choi, Kwanghee and
          Fukayama, Satoru and Shim, Hye-jin and Deshmukh, Soham and Watanabe, Shinji},
  booktitle={WASPAA}, year={2025}
}
```