Training Data

[HuongNV96]

Hi all,
Now,I have three questions,
1.Where I can get training data?
2.noise and clean voice , I need to them to train for model, am not?
3.How can I train for model ? (include steps, preprocess, test , train)
Best Regards,
HuongNV

[ksathyan]

(1) I am not sure about original training dataset which was used. (2) However, you can refer to Microsoft's paper (https://www.isca-speech.org/archive/Interspeech_2019/pdfs/3087.pdf) and github repo (https://github.com/microsoft/MS-SNSD) for creating a training dataset with audio samples. (3) For training please see instructions in TRAINING file (scripts are available in training folder). For bug fixes and enhanced demo program which uses a model, refer to pull request #122

[jiangjianping]

@ksathyan ,
How to use MS-SNSD to train rnn?

[redasamik]

في ثلاثاء، ٩ شباط، ٢٠٢١ في ٤:٥٣ ص، كتب jiangjianping < notifications@github.com>:

…

@ksathyan <https://github.com/ksathyan> , How to use MS-SNSD to train rnn? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#116 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AOW7T7D72ZYUMTK5JX6X66TS6CIP3ANCNFSM4LU4CQGQ> .