During one of Rohit's trips to India in 2014, he witnessed one of the largest outbreaks of malaria in Indian history. When thinking of a project and looking at datasets, he thought of this experience and decided that detecting malaria would be a good problem to solve.

What it does

We made an on the go Android app and website that uses machine learning models to diagnose malaria at 95% accuracy.

How we built it

We used Keras to train a machine learning model that predicted malaria based on a dataset of microscopic images provided by the US National Library of Medicine ( We used 4 convolutional layers with max pooling and dropout. We achieved a 95% accuracy with little overfitting and converging loss and used Flask and Tensorflow Lite to serve these models to the web and Android respectively.

Challenges we ran into

We had issues making the Python model work with Flask and Tensorflow Lite. None of us knew either package before so there were a lot of errors that we spent hours fixing.

Accomplishments that we're proud of

We are proud that we made a software that can potentially help millions of people affected by malaria in developing countries. There are microscopes called FoldScopes that can be added to a phone for just $1 making this app/website very accessible.

What we learned

We learnt a lot about machine learning, convolutional neural networks, and network layers using Keras. We learnt about how truly devastating maleria is. We also learnt how to integrate Python on the web and Android using Flask and Tensorflow Lite respectively.

What's next for MALEX

We want to train our models for more epochs and with more layers. We also want to add data augmentation such as rotations or flips. We were limited by the time and resources of a hackathon, but we want our models to be as accurate as possible. We want to create an IOS app and improve on our Android and web interface since having more platforms makes our product more accessible for everyone.

Built With

Share this project: