lukashaas / cs329s-covid-prediction

This repository is a work-in-progress ML production pipeline for detecting Covid-19 from cough sounds and clinical information.

Covid-19 Prediction App

This app is a Streamlit front end predicting the risk of Covid-19 based on recorded cough sounds. NOTE: This app is purely a proof of concept with no guarantees or warranties and results should not be relied on in any medical setting.

Demo

Current Deployment - Cloud Run

We currently have the app deployed on GCP Cloud Run. For the older deployment version (which is not reliable), please see the App Engine deployment.

Scan the QR code below or click the link to see the app in action:

Covid-19 Risk Evaluation App - Cloud Run

Old Deployment - App Engine

The below is the old deployment of our app which will soon be removed:

Scan the QR code below or click the link to see the app in action:

Covid-19 Risk Evaluation App - App Engine

Project Setup

Dependencies

It is recommended to first install all required dependencies using a virtual environment.

Go to the project directory to which you have downloaded all of this repo's files and run:

cd my-project/
pip install virtualenv
virtualenv venv

Now activate the newly created virtual environment:

source venv/bin/activate

Next, you need to install all of the project's dependencies:

pip install -r requirements.txt

Environment Variables

In order not to expose GCP information to the public, you will need to set some environment variables.

Please create a file called app_environment.sh in the project directory and add the following lines to it, replacing the right side of the equal signs with your GCP project details:

export GCP_PROJECT=your_project
export GCP_REGION=your_region
export GCP_COUGH_STORAGE=your_storage
export GCP_MODEL=your_model

Next run the shell file using source in order to set the environment variables:

source ./app_environment.sh

Running the Application

Run the following command from your project directory to start the Streamlit app:

streamlit run app.py

This will run the application on your local machine.

Cough sounds will not be saved to any device but are processed using GCP APIs.

Deployment

In order to deploy the application to Google Cloud's App Engine, run the following command from the root directory of the project:

make gcloud-deploy-flex

This will create a new app version in a "flexible" GCP App Engine environment.

System Architecture

MVP

The below diagram illustrates how our system on looked like on Google Cloud Platform when we submitted it for MVP evaluation:

V1: CS 329S Demo

Since our MVP, we added a lot of features. The below diagram illustrates our current system architecture on Google Cloud Platform: