반응형
[Paper 리뷰] High-Fidelity Audio Compression with Improved RVQGAN
High-Fidelity Audio Compression with Improved RVQGANLanguage model의 핵심 component는 high-dimensional natural signal을 low-dimensional discrete token으로 compress 하는 neural codec임Improved RVQGANAdversarial, reconstruction loss와 vector quantization technique을 도입하여 high-fidelity의 audio compression을 보장추가적으로 speech, environment, music 등의 다양한 domain에 대한 universal compression을 지원논문 (NeruIPS 2023) : Paper Li..
Paper/Neural Codec
2024. 5. 31. 10:17
반응형