search

ZhihaoDU / du2022sond

Speaker overlap-aware Neural Diarization
GNU General Public License v3.0
10 stars 0 forks source link

When will the code be open sourced? #1

Closed ooobsidian closed 1 year ago

ooobsidian commented 1 year ago

Hi Du, This is a fantastic work! I hope you can open source code, and I will be very happy to learn SOND.

ZhihaoDU commented 1 year ago

Thanks for your attention. Currently, we are making efforts on passing the internal code review. Once it is completed, we will release the code and pre-trained models on https://modelscope.cn

ooobsidian commented 1 year ago

Thank you for your reply. By the way, when is the internal code review expected to be completed?

lavendery commented 1 year ago

I also think it's an excellent work and hope you can open source code as soon as possible.

ZhihaoDU commented 1 year ago

Good news !! Our code will release by March with FunASR repository (https://github.com/alibaba-damo-academy/FunASR), and the pre-trained diarization model with release at ModelScope(https://www.modelscope.cn/models) at the same time. By the way, the pre-trained speaker embedding model has already released at https://www.modelscope.cn/models/damo/speech_xvector_sv-zh-cn-cnceleb-16k-spk3465-pytorch/summary.

ZhihaoDU commented 1 year ago

@ooobsidian @lavendery The inference pipeline has been released in FunASR repository (https://github.com/alibaba-damo-academy/FunASR) including model structures and forward function. The training related codes are under developement. You can read FunASR/egs/alimeeting/diarization/sond/README.md for more details.