Open RefiPeretz opened 6 years ago
I got same results, I put an audio file with only 1 speaker talking and set the n_speakers to 0 for unknown. The result differs every time I try it. Sometimes it recognizes 5, sometimes 6 or 7 speakers,
Any update on how to fix the accuracy?
Hi, i'm facing the same issue. Did you find out a way to implement it for one speaker?
It seems that when use speaker diarization with unknown number of speakers(parameter 0). The script doesn't even consider a case which there is only one speaker in that segment. I can understand that form the way the script define sRange = range(2,10) which means it only going to look from two speakers and above. Am I missing something? is there any approach I can take if I want to perform diarization with now knowing before that I have at least two speakers.