Bird data preprocessing

[haydensflee]

Hello thanks for the interesting github repo. Very cool stuff.

I noticed that you used the bird data from Tchernichovski and Jarvis 2021. I'm trying to find it now.
Did you have to do any data preprocessing? e.g. change sample rate/duration/channel etc?
Thanks.

[GaganNarula]

Hi Hayden Thanks for your interest. I haven't really maintained this repo and now i'm slowly trying to work on it again to get it into a better shape. For your question: I did downsample to 16 KHz because that's easier to train the models with. Single channel instead of stereo. I cropped out silence except for , I believe 200 ms around a sound snippet like a call. There is a lot of silence and noise in the Tchernichovski and Jarvis 2021 data. Do you work on birdsongs as well ? Which lab ? -gagan

…

On Wed, 31 Jan 2024 at 03:53, Hayden Lee ***@***.***> wrote: Hello thanks for the interesting github repo. Very cool stuff. I noticed that you used the bird data from Tchernichovski and Jarvis 2021. I'm trying to find it now. Did you have to do any data preprocessing? e.g. change sample rate/duration/channel etc? Thanks. — Reply to this email directly, view it on GitHub <#1>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACXA42T5DHXT4ZPLQRFDYM3YRGW3PAVCNFSM6AAAAABCSJRPJWVHI2DSMVQWIX3LMV43ASLTON2WKOZSGEYDSMJYGEZDAOI> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[haydensflee]

Hi Gagan,

Thanks for your response. Apologies for late reply. I've been busy and just recovered from being sick as well.
I'm not working in a lab! I'm just a student interested in generative ai, and birdsong synthesis seemed like a cool application since there's already so much TTS out there.
In my studies I've only done some image CNN stuff so RNN with audio is very new to me. Which python notebook is the best working one right now?

Thanks