Skip to content

Latest commit

 

History

History
16 lines (11 loc) · 841 Bytes

README.md

File metadata and controls

16 lines (11 loc) · 841 Bytes

Spkr_Verif_ResNet

In this repository, we used pre-trained Resnet on the VoxBlink2 dataset for speaker verification.

How to do

To run the code:

  1. The required packages are listed in the "requirements.txt" file and you can easily install all of them using: pip install -r requirements.txt

It would be better to make a new Python environment using python3 -m venv myenv , after that, activate the venv using source myenv/bin/activate and then install the packages.

  1. download the pretrained model and copy it in the ./conf/resnet293/model_ft.pt
  2. put your dataset in cv-corpus-17.0-2024-03-15 folder.
  3. open the cv_spkr_clean.py file and change the lang and location of 'tsv files' for that specific lang.