search

pyannote / pyannote-metrics

A toolkit for reproducible evaluation, diagnostic, and error analysis of speaker diarization systems
http://pyannote.github.io/pyannote-metrics
MIT License
186 stars 33 forks source link

Implemented a new metric called SegmentationPurityCoverageFMeasure. #23

Closed diego-fustes closed 5 years ago

diego-fustes commented 5 years ago

The intention of this new metric is to serve as the same purpose as the diarization counter-part, but in this case for segmentation jobs such as Speaker Change Detection

diego-fustes commented 5 years ago

Yes, good point, we can reuse the same coocurrence matrix which is more efficient. All good from my side

hbredin commented 5 years ago

Thanks.

Do you plan to add other stuff to pyannote.metrics in the coming days? If not, I will prepare a new release 1.8 with your latest addition.

diego-fustes commented 5 years ago

No, you can proceed with the release.

El mié., 7 nov. 2018 a las 13:58, Hervé BREDIN (notifications@github.com) escribió:

Thanks.

Do you plan to add other stuff to pyannote.metrics in the coming days? If not, I will prepare a new release 1.8 with your latest addition.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/pyannote/pyannote-metrics/pull/23#issuecomment-436614629, or mute the thread https://github.com/notifications/unsubscribe-auth/AFcV4UnrfhMcFeX_cfULpeK_rae5XtLAks5ustkIgaJpZM4YRBLz .

hbredin commented 5 years ago

I just released pyannote.metrics 1.8.