COVIDCatcher: Developing A Low-Cost Multimodal Machine- Learning Based App for Detecting COVID-19 Symptoms

The Project

I developed COVID-Catcher: a multimodal, low-cost, machine learning based app that can detect COVID-19 symptoms. For symptom detection, several machine learning models were built and compared based on accuracy, recall, and precision using a dataset of 2.7 million patients and their symptoms, with XGBoost coming out on top. For cough detection, a training dataset of ~1445 coughs was processed and used. top performing models were selected for use in COVIDCatcher, in which COVID-19 symptoms are detected using XGBoost, and COVID-19 coughs are identified by a spectrogram, VGG, and a support vector machine. To make these models accessible to the public, I built a web app and deployed both models for users to check for COVID-19 symptoms and learn about COVID-19 by inputting symptoms. The website is accessible on both phones and computers at


About the team

  • United States

Team members

  • Michael