Overlay Augmentation: noise.csv format?

Hi everybody,

im trying to train my model with reallife noise to deploy it later on a traktor.

I want to do that by using the Overlay-Augmentation function of DeepSpeech.py like this:

--augment overlay[p=1,source=/home/...../noise.csv,layers=1:1,snr=40:20~9]

my noise.csv file looks like this:

wav_filename,wav_filesize
01.wav,103450
02.wav,240560

But i have no clue if thats the correct format. I couldnt find any information about the proper setup of the csv format. Can anyone help me out?

thanks alot in advance

@lissyx @reuben , do you guys know anything?
Thanks alot in advance!

We are not working on that anymore, sorry.