Issue in distance threshold parameter - Githubissues

pyannote / pyannote-audio

Neural building blocks for speaker diarization: speech activity detection, speaker change detection, overlapped speech detection, speaker embedding

http://pyannote.github.io

MIT License

6.41k stars 789 forks source link

Issue in distance threshold parameter #1746

Open manish-kumar-iisc opened 4 months ago

manish-kumar-iisc commented 4 months ago

Tested versions

Reproducible in 3.1

System information

pyannote.audio 3.1

Issue description

I was optimizing the clustering threshold, but there was unexpected behavior,

number of speaker clusters: 2, distance threshold: 0.95, keeping rest parameters default
number of speaker clusters: 2, distance threshold: 0.1, keeping rest parameters default
number of speaker clusters: 6, distance threshold: default(0.7), keeping rest parameters default

what can be the reason for the different decrease in the number of speaker clusters on increasing(0.95) & decreasing(0.1)? @hbredin

Minimal reproduction example (MRE)

https://www.youtube.com/watch?v=YoFoKZ9HgfQ&list=PL4x7Of-X4XhgNBVfFpd1N4cSU_X1x96gU&index=2