The second version of soundmerge - customizable benchmark creator for background music detection models.
I made this because there are literally no good labeled music datasets for testing detection/sourcing models
Added new features like support for multiple source directories, choosing between different distributions, updated docs, and made the architecture more reusable. Created smart basic benchmark.
Overall soundmerge is super ready to use, can't wait to finish my models and use the generated benchmark.
github.com/XEas/soundmerge
First version of soundmerge - customizable benchmark creator for background music detection models.
I made this because there are literally no good labeled music datasets for testing detection/sourcing models.
next steps are to fix project structure(so it's not running from src/) and maybe rethink architecture so it can be easily adjusted to different needs
Just generated a 1000-file directory and will be testing my models soon:saluting_face:github.com/XEas/soundmerge