search

taylorlu / Speaker-Diarization

speaker diarization by uis-rnn and speaker embedding by vgg-speaker-recognition
Apache License 2.0
455 stars 124 forks source link

About final output #27

Closed xiaozhi2015 closed 4 years ago

xiaozhi2015 commented 4 years ago

image

Could every output label match the target people of openSLR?

xiaozhi2015 commented 4 years ago

@taylorlu If I build a lib of target people, can I match the output label with my lib?

taylorlu commented 4 years ago

No, the index of the output speaker is start of zero, and the diarization cannot recognize the origin speaker since the embeddings of each sliding window become more continuous rather than discrete as the training speakers.

xiaozhi2015 commented 4 years ago

OK，thx!