Snowdar / asv-subtools

An Open Source Tools for Speaker Recognition
Apache License 2.0
597 stars 135 forks source link

python版本加载PLDA模型打分脚本有提供吗?谢谢 #26

Closed JuneRen closed 3 years ago

1017549629 commented 3 years ago

Hi, actually, in the python version of PLDA, we added the 'plda_trans_write' (https://github.com/Snowdar/asv-subtools/blob/5197a28da4e01ca30f6dd78e1ea50a94b0a96568/score/pyplda/plda_base.py#L210) function to convert the trained PLDA model into Kaldi format, so you can use the Kaldi command 'ivector-plda-scoring' for scoring (see one of the examples https://github.com/Snowdar/asv-subtools/blob/5197a28da4e01ca30f6dd78e1ea50a94b0a96568/score/pyplda/test_PLDA.sh).

JuneRen commented 3 years ago

Hi, actually, in the python version of PLDA, we added the 'plda_trans_write' (

https://github.com/Snowdar/asv-subtools/blob/5197a28da4e01ca30f6dd78e1ea50a94b0a96568/score/pyplda/plda_base.py#L210

) function to convert the trained PLDA model into Kaldi format, so you can use the Kaldi command 'ivector-plda-scoring' for scoring (see one of the examples https://github.com/Snowdar/asv-subtools/blob/5197a28da4e01ca30f6dd78e1ea50a94b0a96568/score/pyplda/test_PLDA.sh).

谢谢您的回复,项目里暂时没有python加载PLDA模型进行打分的脚本吗?

1017549629 commented 3 years ago

Thanks for the reminder. Yes, there is a python script for loading PLDA models for scoring that will be updated at (https://github.com/Snowdar/asv-subtools/blob/master/score/pyplda/gaussian-plda-scoring.py). Please let us know if there are any problems in using the script.

JuneRen commented 3 years ago

Thanks for the reminder. Yes, there is a python script for loading PLDA models for scoring that will be updated (https://github.com/Snowdar/asv-subtools/blob/5197a28da4e01ca30f6dd78e1ea50a94b0a96568/score/pyplda/gaussian-plda-scoring.py). Please let us know if there are any problems in using the script.

Thank you very much for your reply. When will this script be updated? This link is still empty.

1017549629 commented 3 years ago

Now is ok.

JuneRen commented 3 years ago

Now is ok.

Thank you very much