End-to-end-ASR-Pytorch icon indicating copy to clipboard operation
End-to-end-ASR-Pytorch copied to clipboard

Published 20 hours ago • verified publisher icon Alexander-H-Liu

Is there anybody uses custom data using librosa for traning LAS

Open qute012 opened this issue 4 years ago • 0 comments

Hi. I want to change dataset and recognize in korean. I use librosa mel-spectro as extraction features. It shows below. image I didn't modify model parameters and train parameters. But model predicts same class almost one class when i gave some samples.

Below, When i gave sample by using torch.randn, results image It was almost just 2 classes. Is there problems at preprocessing audio file(e.g. mfcc, mel spectro) or models?

anybody knows?

qute012 avatar Feb 28 '20 12:02 qute012