search

sp-uhh / sgmse

Score-based Generative Models (Diffusion Models) for Speech Enhancement and Dereverberation
MIT License
521 stars 76 forks source link

Debugging steps #61

Open libai2345 opened 1 week ago

libai2345 commented 1 week ago

运行结果

The dataset I used was VoiceBank-DEMAND, why was my pesq so low, trained for 35 rounds.