With @ravencheuk we would like to share our recent work on multi-track music transcription, which attempts to improve upon MT3!
MR-MT3: Memory Retaining Multi-Track Music Transcription to Mitigate Instrument Leakage
paper: https://t.co/xBcI7iuRyk
code: https://t.co/Zvy9MI6vSR
I asked ChatGPT to proofread my Japanese writing. I am amazed by the fact that that it could spot out the factual mistake I made due to typo (2014 instead of 2004).
It's actually smarter than I thought.
A couple weeks ago when I first found his channel, he got only a few K subs.
Now, it's approaching 100K subs.
YouTube algorithm is indeed interesting.
https://t.co/QuZ4L4fSLU
``Moisesdb: A dataset for source separation beyond 4-stems. (arXiv:2307.15913v1 [https://t.co/mPAjnto8C8]),'' Igor Pereira, Felipe Araújo, Filip Korzeniowski, Richard Vogl, https://t.co/nyvTpDGkVv
There are so many hidden gems on GitHub.
Another invertible CQT implementation, a much more concise one.
I wish I had a larger number of followers so that I could bring these hidden gems to light.
https://t.co/lL42ujaIym