We initially wanted to help take off the load from healthcare workers by using neural networks to make it easier to diagnose COV1D-19 from CT scans of lungs. While we were working on Corona Bay, we realized that we could also help the general public by giving them the information and resources to understand the situation of the pandemic and self-diagnose for coronavirus.

What it does

Corona Bay has 4 main features.

1) Detection: a neural network model we created and trained from scratch that predicts if a patient has COVID-19 based on CT scans of their lungs. This was created and trained using Tensorflow and has 98.09% test accuracy.

2) Statistics: a dynamic map that lets users hover over each country to get the stats of the pandemic situation in all of the countries. This information is always updating and can help people understand where the situation is worse and how they might need to act based on the situation in their country.

3) News: a constantly updating feed of coronavirus-related news. This allows users to stay up-to-date on the latest information about the pandemic while staying on one site.

4) Symptoms: a checklist of symptoms that allows users to get a good idea what they need to look out for and what actions they need to take if they have any of those symptoms.

How I built it

We used React.js for most of the project, with Python 3 and Tensorflow Keras API for the machine learning model.

Challenges I ran into

Integrating Tensorflow in Python to the web app created so many issues. It was really hard to just migrate to JavaScript, and then we got memory leaks when trying to use it. Also, React.js doesn't make it easy to style components or have an overall layout, so it was hard to make the UI presentable. The datasource we used for the statistics page actually went down right when we were recording the video! It took a lot of effort and determination to conquer these challenges.

Accomplishments that I'm proud of

We are very proud that we were able to get all 4 features working well and that we were able to finally get rid of the errors. The ML model worked surprisingly well with a small dataset of less than 1000 images. We are also very proud that the interactive map is easy to use and visually shows the effect of the pandemic.

What I learned

We learned how to use machine learning to solve real-world problems with limited resources. We also learned how to use React.js to make a fast web app that can keep up with users' needs.

What's next for Corona Bay

We only had a short time to develop this project, so we didn't have the time to style everything well. We were mainly focused on getting the best features out there, which was especially hard due to the incredible number of errors that came up in the process. We are looking to shape the web app so that it can be used by the public and refine the minor styling details so that we can present our great features in the best way.

Try It out



mongodb, react.js, tensorflow

Devpost Software Identifier