This is an open source project (formerly named Listen, Attend and Spell - PyTorch Implementation) for end-to-end ASR implemented with Pytorch, the well known deep learning toolkit.
MIT License
1.18k
stars
317
forks
source link
Is there anybody uses custom data using librosa for traning LAS #55
Hi. I want to change dataset and recognize in korean.
I use librosa mel-spectro as extraction features. It shows below.
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
It was almost just 2 classes. Is there problems at preprocessing audio file(e.g. mfcc, mel spectro) or models?
Hi. I want to change dataset and recognize in korean. I use librosa mel-spectro as extraction features. It shows below.
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
It was almost just 2 classes. Is there problems at preprocessing audio file(e.g. mfcc, mel spectro) or models?
anybody knows?