Skip to content

Unable to Reproduce Default Results on VoxConverse Dataset #6

@hongjiaying

Description

@hongjiaying

Description:

Dear Author,
Hello! I am currently using the SimpleDiarization project for speaker diarization tasks and attempting to reproduce the results you provided on the VoxConverse dataset. I followed the installation and configuration instructions in the README file, but my results differ from the default results you provided. I would like to ask if there are any specific configurations or dependency versions that I should pay special attention to, or if I have missed any steps.

My Environment and Steps:

Operating System: macOS
Python Version: 3.11

Dependency Versions:
pyannote.audio==3.3.2
speechbrain==1.0.3
scikit-learn==1.7.1
torch==2.7.1
tqdm==4.67.1
pyyaml==6.0.2

My Results:

Development Set (dev):
OVERALL DER : 5.90% JER : 17.17%
Test Set (test):
OVERALL DER : 8.43% JER : 25.04%

My Questions:

  1. My results differ from the default results you provided, especially the JER value for the test set. Have I missed any steps or do I need to adjust certain configurations?
  2. Are there any other suggestions or considerations that could help me reproduce your results?

Thank you very much for your help and support!

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions