Inspiration

In Covid times when the doctors are stretched too thin, our preliminary diagnostic tool serves to quickly detect whether a chest xray belongs to a healthy person or to someone who has developed pneumonia.

What it does

It can be used easily by medical and non-medical persons alike. Pneum-o-ray accurately detects pneumonia 91% of the times.

How we built it

We trained a neural network model (VGG16) using keras on a Kaggle dataset of chest x-rays. We then built an API to host our model. We simultaneously built a front-end and connected the same to our API.

Challenges We ran into

We had some issues in uploading images to API and passing them on to the model. That took some debugging.

Accomplishments that We're proud of

We had a team with diverse skillsets. Ketki was well versed with Keras, Justin and Nandan were skilled at front-end and design, and Vlad was proficient with Django. It was amazing how we each were able to communicate our requirements and make it work.

What we learned

Besides from the fact that we were all exposed to technologies we hadn't worked on before, we were relatively new hackers, and we were amazed at what we could achieve.

What's next for Pneum-o-ray

We could work on achieving better accuracy. We are at 91% currently, which is not bad, but we could push that number higher. We also had plans to implement a google maps API on our results page to display doctor recommendations, it would be amazing to implement that functionality.

Built With

Share this project:

Updates