Stuck on an issue?
Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.
And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.
Two step with single source
See original GitHub issue❓ Questions and Help
Maybe this is obvious to some people, but to me it wasn’t. The filterbank training part of the two step paper does not seem to work when you have a single source only. It does not converge anywhere useful; stays around 0 dB SI-SDR. As a workaround, using (mixture, (source, mixture - source)) as training x, y pairs seems to work OK.
Issue Analytics
- State:
- Created 3 years ago
- Comments:9 (1 by maintainers)
Top Results From Across the Web
How to Two Step: 14 Steps (with Pictures) - wikiHow
Start by getting into your dancing position, with one partner leading. Then, practice moving in a quick, quick, slow, slow pattern to a...
Read more >Country-western two-step - Wikipedia
The Texas two-step is the same step known to ballroom dancers as the international fox-trot. Except for the one-step, which is just that,...
Read more >Two-step equations review (article) - Khan Academy
A two-step equation is an algebraic equation you can solve in two steps. Once you've solved it, you've found the value of the...
Read more >History of the Two Step | Country Dancing - Grizzly Rose
The dance is often called the Texas Two Step or the Country Two Step, and those names came from the period when it...
Read more >Part 6 - Competition Requirements | Acquisition.GOV
6.302-1 Only one responsible source and no other supplies or services will ... of competitive procedures (e.g., two-step sealed bidding).
Read more >
Top Related Medium Post
No results found
Top Related StackOverflow Question
No results found
Troubleshoot Live Code
Lightrun enables developers to add logs, metrics and snapshots to live code - no restarts or redeploys required.
Start Free
Top Related Reddit Thread
No results found
Top Related Hackernoon Post
No results found
Top Related Tweet
No results found
Top Related Dev.to Post
No results found
Top Related Hashnode Post
No results found
The two-step approach tries to optimize the target masks on the latent space where the input mixture is composes from at least 2 sources. So you need to provide the clean sources (or source and noise if it is denoising) in order to find the optimal target. So it is totally expected to have 0db if you only provide only one source that consists to the mixture. Check figure 1 in the paper https://arxiv.org/pdf/1910.09804.pdf
Thanks for your answer Thymios. I’ll close !