how about the result?

[wuqiangch]

how about the result?

[johnkorn]

well, I got 97%/93% accuracy on train/validation sets, but I did not launch training on full LibriSpeech dataset, only on clean audios, so the model is not ready for real life usage yet. Thanks for the interest though.

[ovninosa]

@johnkorn there is a tutorial to run your proyect?

[wuqiangch]

@johnkorn Have you read the paper "Deep Speaker: an End-to-End Neural Speaker Embedding System"?Maybe the method is the best one of speaker recognition.

[johnkorn]

@javenosa no, I did not finish the work on the project because had to switch to other tasks, so no tutorial yet.

@wuqiangch thanks for the paper, looks like a good approach, exactly what I needed. I'll try to use it when come back to the project.