Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

GUI: advanced option to enable training-time upsampling so that it is compatible with all training WAVs #354

Closed
wants to merge 5 commits into from

Conversation

honkkis
Copy link
Contributor

@honkkis honkkis commented Jan 7, 2024

Hi, I think I found a good way to allow antialiasing geeks like me to do training-time upsampling. The key idea is to do the upsampling in memory after the wavs have been loaded, validated, and split.

This pull request fixes #355 .

Please merge to main branch or let me know if something needs to be fixed to be compliant to main. Unit tests pass.

Screenshot 2024-01-07 at 17 50 05

Benefits shown in these spectrograms:

alias

@honkkis honkkis changed the title GUI: advanced option to do during-training upsampling so that it is compatible with all training WAVs GUI: advanced option to enable training-time upsampling so that it is compatible with all training WAVs Jan 7, 2024
@honkkis
Copy link
Contributor Author

honkkis commented Jan 7, 2024

Reverted back to using scipy's resample since that shows very clean spectrogram with no content above nyquist 24kHz after resampling to 96 kHz.
Screenshot 2024-01-08 at 0 47 52

This was not the case with linear upsampling.

@sdatkinson
Copy link
Owner

Thanks for the PR.

Like I said in my response in #355, this isn't the direction I'm currently looking to take NAM's standardized trainer, but it's a cool piece of work nonetheless 🙂

@sdatkinson sdatkinson closed this Jan 13, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

[FEATURE] GUI: Add advanced option to do training-time upsampling
3 participants