Audio Annotation Visualizer

Simple flask server that uses wavesurfer.js to visualize and check for quality of annotations of your audio data. The interface looks like the image shown below.

Setup

1. Clone the Repository: Clone the repository to your local machine using the following command:

   git clone git@github.com:WadhwaniAI/audio-annotation-visualizer.git
   cd audio-annotation-visualizer

2. Create and activate a Conda Environment: Run the following command to create a conda environment named visualizer using Python 3.10:

   conda create -n visualizer python=3.10

   conda activate visualizer

3. Install Dependencies: Use the requirements.txt file to install all necessary dependencies:

   pip install -r requirements.txt

4. Set Environment Variables: Copy the .env.example file to .env and set the following environment variables:

   USERNAME=<username>
   PASSWORD=<password>
   AUDIO_ROOT_DIR=<path_to_audio_data>

5. Run the Server:

   python main.py

Data Directory Setup

Set audio_root_dir in the .env file. The structure of the audio_root_dir should look like this:

• This is the path where all the audio data, corresponding reference text and annotation files are stored. Inside this directory:
• Each subdirectory represents a single data sample.
• Each subdirectory contains:
  • audio.mp3: A recording of the sample.
  • ref_text.txt: (optional) A text file containing the reference text of the audio.mp3 recording. This is relevant for audios captured in a read aloud setting where the speaker is given a text to read.
  • annotation.json: A json file containing annotations corresponding to the audio.mp3 recording.

Example structure:

audio_root_dir/
├── sample_1/
│   ├── audio.mp3
│   └── ref_text.txt
│   └── annotation.json

Third-Party Library Notices

This project uses wavesurfer.js, an open source audio waveform visualization library.

wavesurfer.js is licensed under the BSD-3-Clause License.

Copyright (c) 2012-2023, wavesurfer.js contributors. All rights reserved.