search

mravanelli / pytorch-kaldi

pytorch-kaldi is a project for developing state-of-the-art DNN/RNN hybrid speech recognition systems. The DNN part is managed by pytorch, while feature extraction, label computation, and decoding are performed with the kaldi toolkit.
2.36k stars 446 forks source link

voxceleb cfg #86

Closed hdubey closed 5 years ago

hdubey commented 5 years ago

Hi Mirco, Thanks for this great toolkit. I am trying to re-produce some of the experiments from MI training of SincNet paper Here:https://arxiv.org/pdf/1812.00271.pdf Is the related codes for voxCeleb available in this repo. I am interested it is as it is noiser and realistic as compared to TIMIT or Librispeech. Thanks!

mravanelli commented 5 years ago

It is not currently available, but we are working to extend this repository to other tasks such as speaker recognition. In the next months, you will probably find also the mutual-information approach proposed in that paper. Stay tuned!

tombarti commented 5 years ago

Hi Mirco, First of all thank you for your fantastic research! I too am very interested in reproducing some of the experiments from your paper "Learning Speaker Representations with Mutual Information", especially for speaker diarisation. Is there any update on adding speaker recognition (and in particular speaker verification) to this repository? Thanks!

mravanelli commented 5 years ago

Hi Tom, the goal is to drastically extend the current version of pytorch-kaldi in many different ways (including adding LIM or other self-supervised techniques such as PASE (https://arxiv.org/abs/1904.03416)). The project will get bigger and we already have some sponsors for that. At the beginning of September there will be an announcement for that. Stay tuned!

Mirco

On Mon, 19 Aug 2019 at 11:59, Tom Bartissol notifications@github.com wrote:

Hi Mirco, First of all thank you for your fantastic research! I too am very interested in reproducing some of the experiments from your paper "Learning Speaker Representations with Mutual Information", especially for speaker diarisation. Is there any update on adding speaker recognition (and in particular speaker verification) to this repository? Thanks!

— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub https://github.com/mravanelli/pytorch-kaldi/issues/86?email_source=notifications&email_token=AEA2ZVSGLY5XTWEYQED3NYDQFK7MPA5CNFSM4HIJIWE2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4TOAWA#issuecomment-522641496, or mute the thread https://github.com/notifications/unsubscribe-auth/AEA2ZVRWG3AQJUGXYM7ZDADQFK7MPANCNFSM4HIJIWEQ .

tombarti commented 5 years ago

Thank you for the swift reply Mirco, it is much appreciated! I am looking forward to the September announcement, especially LIM and PASE!

LCF2764 commented 4 years ago

Hi Mirco, The literature "Learning Speaker Representations with Mutual Information" says that the code is open source. I haven't found the relevant code here. Is it open source now? Thanks!

mravanelli commented 4 years ago

Hi, the plan is to put it in the speechbrain project directly https://speechbrain.github.io/

On Apr 27, 2020 05:11, "LCF2764" notifications@github.com wrote:

Hi Mirco, The literature "Learning Speaker Representations with Mutual Information" says that the code is open source. I haven't found the relevant code here. Is it open source now? Trhanks!

— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub https://github.com/mravanelli/pytorch-kaldi/issues/86#issuecomment-619844736, or unsubscribe https://github.com/notifications/unsubscribe-auth/AEA2ZVTOEAKFNMBRFDLAIKTROVD5JANCNFSM4HIJIWEQ .

Soar-Sir commented 4 years ago

Hi Mirco, as a developer in the field of speech, I am very interested in SpeechBrain. So when will the public version of the code be released? I am looking forward to this project. Thanks a lot!

TParcollet commented 4 years ago

Dear Soar-Sir,

Thank you for your interest in SpeechBrain. Due to the very complex COVID situation, we had to postpone a little bit the release. Right now, an expected period is Fall 2020.