search

deezer / spleeter

Deezer source separation library including pretrained models.
https://research.deezer.com/projects/spleeter.html
MIT License
25.76k stars 2.82k forks source link

Full signal reservation #481

Open ItId123 opened 4 years ago

ItId123 commented 4 years ago

Hi,

I would like to understand, when Spleeter separates a track to 4 STEMS, for example, (with 4 masks I guess) how do it makes sure that no data is missing? I mean, is it compares the sum of the 4 output tracks to the original one? Or, maybe there are masks only for the first 3 STEMs , and the rest goes to the 'Others'?

Thanks

brisu commented 3 years ago

I believe it would be applying the first 3 masks and the other data would be assigned to 'Others'