Speech Audio Annotation System [WIP]

This tool uses Mozilla DeepSpeech as an example to demonstrate the some generic functionalities / components of a speech-audio annotation tool, including annotation and model inference - training loop, to semi-automate the annotation process.

This repo contains the back-end component of the annotation system, the front-end component speech-audio-annotation-ui is also required for the usage of the tool.

Setup Back-end Component

1. Clone the project git repo

2. Download and unzip the pre-trained DeepSpeech model to <project_root_dir>/outputs:

cd <project_root_dir>/outputs
curl -LO https://github.com/mozilla/DeepSpeech/releases/download/v0.6.1/deepspeech-0.6.1-models.tar.gz
tar xvf deepspeech-0.6.1-models.tar.gz

3. Clone the DeepSpeech repo to <project_root_dir>/models/deepspeech (only required if model-training is needed):

git clone https://github.com/mozilla/DeepSpeech.git <project_root_dir>/models/deepspeech`

4. Build and run docker-compose in <project_root_dir> to bring up the containers:

docker-compose up --build

Setup Front-end Component

1. Clone the UI project repo outside the back-end project repo:

git clone https://github.com/francesliang/speech-audio-annotation-ui

2. Build and run docker-compose in <project_root_dir> to bring up the containers:

docker-compose up --build

3. The URL of the UI should be localhost:3000

How it works