aim-qmul / sdx23-aimless

Source Separation training codebase for the Sound Demixing Challenge 2023.
37 stars 0 forks source link

implement DCUNet #7

Open yoyolicoris opened 1 year ago

csteinmetz1 commented 1 year ago

Started on this, but seems to be non-trivial to get it to train. Lots of patching of Asteroid is required for MSS application. Saurjya thinks there may not be a compelling reason to use DCUNet for MSS, so will hold on implementing this further.