search

facebookresearch / svoice

We provide a PyTorch implementation of the paper Voice Separation with an Unknown Number of Multiple Speakers In which, we present a new method for separating a mixed audio sequence, in which multiple voices speak simultaneously. The new method employs gated neural networks that are trained to separate the voices at multiple processing steps, while maintaining the speaker in each output channel fixed. A different model is trained for every number of possible speakers, and the model with the largest number of speakers is employed to select the actual number of speakers in a given sample. Our method greatly outperforms the current state of the art, which, as we show, is not competitive for more than two speakers.
Other
1.23k stars 178 forks source link

make_dataset.py is slow to generate samples #74

Closed srdfjy closed 1 year ago

srdfjy commented 2 years ago

Hi @adiyoss, I used make_dataset.py to generate custom samples, and I also switched to pyrirgen, but it is still very slow (one is generated every second on average), and adding multiple threads has no effect. Is there any way to speed up the generation of samples?

adiyoss commented 2 years ago

Hi @srdfjy, Unfortunately no. You can parallelize this code manually, but we do not have any support for it at the moment.

srdfjy commented 2 years ago

@adiyoss thanks!

srdfjy commented 2 years ago

make_dataset.py is randomly selected when generating wav, and does not consider repetition, so it may generate repeated audio, which will affect the model effect?