Skip to content

COVIDNet is a project that intends to predict COVID-19 infection status solely by vocal audio recordings.

Notifications You must be signed in to change notification settings

aem713/COVIDNet

Repository files navigation

COVID-19 Prediction Model From Audio Data


Project Description

DSCI 400 COVID-19 Audio Project This process is to create a predictive model for COVID-19 from a set of audio files. Broadly, the steps are data cleaning, generating and normalizing Mel-spectrograms and handcrafted features, testing and training a model composed of a CNN and a Perceptron, and validation of the results.


Instructions

  1. Download the CovidNet_Final.ipynb notebook
  2. Open this notebook in Google Colab
  3. Create a shortcut to the 'DSCI 400' Google Drive folder (https://bit.ly/3nUvQyM) in your main Google Drive. The path should be '/content/gdrive/MyDrive/DSCI400'
  4. Run the COVIDNet notebook. All other required APIs are included in the 'DSCI 400' folder provided in the link.

It is important to note that the only file needed to use COVIDNet is the CovidNet_Final.ipynb notebook. All the other files included in this repository are simply for ease of viewing, as they are a replica of the folder system in Google Drive. In the interest of saving memory, any interested user need not download the entire repository; they only need to download the CovidNet_Final.ipynb notebook.

Dependencies

The Libraries required for this program are:

INTERNAL TO PYTHON VERSION 3.9.2

  • os
  • sys
  • subprocess
  • glob

EXTERNAL TO PYTHON VERSION 3.9.2

  • google-colab 1.0.0 (2021)
  • google 0.1.1 (2021)
  • librosa version 0.8.0 (2021)
  • pandas version 1.2.3 (2021)
  • numpy version 1.20.0 (2021)
  • surfboard 0.2.0 (2021)

Dataset

The dataset used for this project is under "Coswara Original Files". The files can also be sourced at "https://github.com/iiscleap/Coswara-Data".

About

COVIDNet is a project that intends to predict COVID-19 infection status solely by vocal audio recordings.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published