In this work, we aim to analyze and optimize the EnCLAP framework, a state-of-the-art model in automated audio captioning. We investigate the impact of modifying the acoustic encoder components, explore pretraining with different dataset scales, and study the effectiveness of a reranking scheme. Through extensive experimentation and quantitative analysis of generated captions, we develop EnCLAP++, an enhanced version that significantly surpasses the original.
@article{arxiv.2409.01201,
title = {EnCLAP++: Analyzing the EnCLAP Framework for Optimizing Automated Audio Captioning Performance},
author = {Jaeyeon Kim and Minjeon Jeon and Jaeyoon Jung and Sang Hoon Woo and Jinjoo Lee},
journal= {arXiv preprint arXiv:2409.01201},
year = {2024}
}