search

sp-uhh / sgmse

Score-based Generative Models (Diffusion Models) for Speech Enhancement and Dereverberation
MIT License
491 stars 74 forks source link

Dereverberation not work #3

Closed loboere closed 2 years ago

loboere commented 2 years ago

I run !python enhancement.py --test_dir "/content/test" --enhanced_dir "/content/na" --ckpt "/content/drive/MyDrive/SGMSE+ Dereverberation Checkpoints/epoch=326-step=408750.ckpt"

but the results are not the same as in the demo page ,the sound comes out saturated

cobalamin commented 2 years ago

Hi, sorry about this. We had accidentally removed some options for the network that we thought were unused. It should be fixed on the most recent commit, https://github.com/sp-uhh/sgmse/commit/c2641a9a0b88341563fa95983e60fe0cf377d795 . I checked both SE and Dereverberation again.

Note that changing the sampler settings slightly seems to work better for dereverberation, using --N 50 --snr 0.33 (I'll also add this info to the README).

Let me know if it works or if something's still broken for you.

cobalamin commented 2 years ago

Closing this now due to inactivity. Feel free to open another issue if something is still broken for you :)