Torch Voice activity detection

This project has been created as a test task for an employment.

Clone

Firstly, clone this repository:

git clone https://github.com/ilyailyash/Torch-Voice-activity-detection.git
cd Torch-Voice-activity-detection

Environment && Installation

Install Anaconda or Miniconda, and then configure the environment with requirements.txt:

# create a conda environment
conda create --name <env> --file requirements.txt
conda activate <env>

Documentation

• Model Description
• Getting Started

Pretrain models:

Model	EER	FPR with FNR = 1%	FNR with FPR = 1%
Model for config/train/vad.toml	9.9%	82%	17%
Model for config/train/vad_360.toml	8.8%	77.8%	13.3%

Presented models are on training and will be updated soon.

Comparator with baseline model

Model	F1
webrtcvad	0.945
Proposed	0.95

License