The whole world has come to a standstill to fight the major pandemic that has had such a massive effect globally. Being a technologist, we would like to solve this problem faced by the world during pandemic outbreak. Testing people in the early phase of outbreak is one of the solution to stop pandemic. To test more people we need to filter people with high chances of infection rather than blindly testing everyone. Our idea aims to solve this issue of effective testing of COVID-19 affected people, to help ease the government and provide effective use of testing kits.

What it does

A preliminary analysis of covid-19 is created based on :

  • AI based dry cough recognition
  • Body temperature using wearable or user input
  • User risk group(age,travel history,health conditions,first contact with infected patients).

A score is generated which presents probability of COVID-19 infection for the application users.

How I built it

  • The core part of the application is the CNN based deep learning algorithm that detects dry cough. The other details input by user are evaluated and a scoring mechanism provides the preliminary score.
  • The model is trained using cough data samples from various sources like YouTube, OSF and personal recordings from friends and relatives.
  • The frontend is built using HTML and Ajax, while the backend is developed using Flask and Python. The server is built on AWS (Amazon Web services) EC2, which provides a secure SSL based HTTPS method to save the audio and user data.

Challenges I ran into

  • Development of Front end and backend, since none of the contributors have experience in this domain. We all are electrical and deep learning engineers with no practical experience in Full stack development.

Accomplishments that I'm proud of

What's next for COVID-19 Preliminary diagnostic application

  • Create a mobile application based on similar features and deploy it on PlayStore.
  • Domain Registration Ongoing :
  • Maintain a private website, that keeps a record of the probability of affected patients. This can be broken down to specific regions, locality, etc.
  • Collaborate with government agencies and provide them with this data so that they can organize testing affected patients and help them fight this issue effectively.

Built With

+ 3 more
Share this project:


posted an update

We have successfully implemented a model that distinguishes cough and not cough from an audio source file. The team is now actively working on a model that classifies normal cough and dry cough. We are still looking out for Android app developers who can help us build this app.

Log in or sign up for Devpost to join the conversation.