Efficient Audio Captioning with Encoder-Level Knowledge Distillation

TL;DR

This paper introduces an encoder-level knowledge distillation framework for audio captioning that improves model efficiency and performance, especially in data-scarce scenarios, by distilling knowledge into the encoder using contrastive loss.

Contribution

It proposes a novel encoder-level knowledge distillation method for AAC, demonstrating improved robustness and efficiency over traditional approaches.

Findings

Contrastive KD outperforms MSE KD in robustness.

Student model achieves 19x faster inference.

Leveraging audio-only data enhances performance.

Abstract

Significant improvement has been achieved in automated audio captioning (AAC) with recent models. However, these models have become increasingly large as their performance is enhanced. In this work, we propose a knowledge distillation (KD) framework for AAC. Our analysis shows that in the encoder-decoder based AAC models, it is more effective to distill knowledge into the encoder as compared with the decoder. To this end, we incorporate encoder-level KD loss into training, in addition to the standard supervised loss and sequence-level KD loss. We investigate two encoder-level KD methods, based on mean squared error (MSE) loss and contrastive loss, respectively. Experimental results demonstrate that contrastive KD is more robust than MSE KD, exhibiting superior performance in data-scarce situations. By leveraging audio-only data into training in the KD framework, our student model achieves competitive performance, with an inference speed that is 19 times faster\footnote{An online demo is available at \url{https://huggingface.co/spaces/wsntxxn/efficient_audio_captioning}}.