반응형
[Paper 리뷰] DFlow: A Generative Model Combining Denoising AutoEncoder and Normalizing Flow for High Fidelity Waveform Generation
DFlow: A Generative Model Combining Denoising AutoEncoder and Normalizing Flow for High Fidelity Waveform GenerationHigh-fidelity의 waveform generation을 위한 vocoder가 필요함DFlow고품질 생성을 위해 Normalizing Flow와 Denoising AutoEncoder를 결합추가적으로 model size와 training set을 확장하여 DFlow를 large-scale universal vocoder로 scaling up논문 (ICML 2024) : Paper Link1. IntroductionDeep Generative Model (DGM)은 waveform generat..
Paper/Vocoder
2024. 7. 7. 13:27
반응형