Downloads · 30 days
0
dereverb_mel_band_roformer
anvuew/dereverb_mel_band_roformer
dereverb_mel_band_roformer is a machine learning model from anvuew. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as gpl-3.0.
专用于人声的去混响模型，适用于 https://github.com/ZFTurbo/Music-Source-Separation-Training
Access
Public
Updated Feb 11, 2025
Repo size
4.9 GB
Likes
23
Public
Repo makeup
Click a slice to open those files.
.ckpt4.9 GB · 100%
At a glance
- License
- gpl-3.0
- Access
- Public
- Created
- Jul 12, 2024
- Updated
- Feb 11, 2025
- SHA
- cef05ad2
Model card
From the Hugging Face model README
专用于人声的去混响模型，适用于 https://github.com/ZFTurbo/Music-Source-Separation-Training
微调自kim的人声乐器分离模型，所以保留了一些分离人声与乐器的能力
训练数据的干声为单声道单人演唱/说话，所以模型会尽可能使分离后的干声为单声道，导致在分离一些人声为立体声或者经过叠轨等处理的音频时可能会出问题，dereverb_mel_band_roformer_less_aggressive_anvuew_sdr_18.8050.ckpt是使用为这些情况的样本挑出来的dereverb_mel_band_roformer_anvuew_sdr_19.1729.ckpt训练中期的权重
dereverb_mel_band_roformer_anvuew_sdr_19.1729.ckpt与dereverb_mel_band_roformer_less_aggressive_anvuew_sdr_18.8050.ckpt两个权重由于训练代码bug，在训练时混响与人声实际并未对齐，所以这两个模型表现出一定程度的去除人声与乐器分离后的残留（通常是弦乐）以及部分和声（非中央声道）的能力
dereverb_mel_band_roformer_mono_anvuew_sdr_20.4029.ckpt则因修复了对齐bug，去混响的能力更强了，但是去人声乐器分离残留以及和声的能力下降了
- License
- gpl-3.0
- Created
- Jul 12, 2024
- Updated
- Feb 11, 2025