What is the form of data directory structure? Can you provide an example?

[WhiteFu]

Hi, great work.
What is the form of data directory structure when the speaker encoder is trained? Can you provide an example?
I want train the model in own dataset.
Thanks for your share.

[alessandropec]

I thank you for appreciate. The speaker encoder is trained using VoxCeleb dataset, so the form of data directory structure is:

root->speakers_folder->video_folders (i will call this subfolders rec_sessions) ->wavs

In this case for each speaker there are several video folders (to handle this specify multi_rec_session=True) inside each rec_session there are the wavs.

If you do not have the rec session folder so the structure is like:

root->speakers_folder->wavs

Set multi_rec_session=False and the training should be run.

I hope I have helped you.