Pleased to announce my first ML-audio paper, now on arXiv: "SignalTrain: Profiling Audio Compressors with Deep Neural Networks" https://arxiv.org/abs/1905.11928 Interactive demo & audio clips: http://www.signaltrain.ml pic.twitter.com/CteA37DONw
-
Thank you very much! Had not considered that. Actually was thinking of going "the other way" -- had a couple students trying to port
@citnaj's DeOldify over to audio, in prep for a colleague's (manual) audio restoration grant -- but adding tape effects? Why not?! :-)
Wow that’s an awesome idea! Definitely curious to see how that goes. I’ve been thinking NoGAN training in particular should be useful for more than just images.
-
-
Follow-up: Looks like StatisticDean had the same idea as my students: i.e., start with the
@fastdotai image-restoration example but for (mono) audio spectrograms (to 3 'colors': real, imag, mag): https://github.com/StatisticDean/Crystal_Clear … But so far no NoGAN (AFAIK), etc, just VGG & feature norm -
Cool. Honestly I’d start there and see how far it goes. NoGAN is easy to add later.
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.