Scripts and utilities for working with the VOiCES dataset
Instructions for downloading and descriptions of the VOiCES dataset can be found here. The code in this repo is designed for the VOiCES_devkit and VOiCES_release.
This repo is divided into several subdirectories, briefly described below. Further details can be found in READMEs of the subdirectories.
This directory contains scripts for building up index files of the data set, or converting those index files into a format compatible with Nvidia NeMo ASR.
This directory contains class definitions for a PyTorch dataset that can be used to train a speaker verification model.
This directory contains class definitions and scripts to facilitate ASR inference on VOiCES data using NeMo and Quartznet.