Inspiration

600 million people contract malaria each year. Of those people, over 1 million die. 90% of those deaths occur in Sub-Saharan Africa, where people don't have easy access to doctors. We want to help people like them determine their risk of contracting a deadly bug-borne disease and evaluate if they should see a doctor. Malaria can almost always be cured if it's detected earlier and treated promptly.

What it does

We use machine learning and the current weather to predict your odds of contracting a deadly bug-borne disease such as malaria. If you have a bug bite, you can take a picture using the app and it will use IBM Watson's Visual Recognition API to detect the type of bug that bit you. Some bug bites, such as tick bites, are relatively harmless. If it'll take a lot of effort to see a doctor, it may not be worth it to go if you have a tick bite.

How we built it

We built a mobile web app with JQuery for the front-end and Express for the back-end. We collected hundreds of photos of bug bites and abnormal skin conditions from Google and Baidu. Then, we uploaded them to IBM Watson to train a custom classifier.

Challenges we ran into

The biggest challenge we went through is we couldn't find enough good-quality images of bug bites. Most images were too small, grainy, or taken from an odd angle. Since there wasn't enough training data, our classifier was very inaccurate. Our solution was to apply transformations to the images we've found to produce new images. We made sure there was enough distortion that the classifier wouldn't be overfitted. Also, this is Lucy's first hackathon, so there's a lot to get used to (but the Red Bull helped!)

Accomplishments that we're proud of

This was the first time either of us developed a machine learning application that isn't contrived. Previously, we've primarily done machine learning for assignments or competitions, where the results aren't very application to real life. Other the other hand, Bite D.tech was able to detect that Leo's bug bite is a mosquito bite!

What we learned

We learned how to use API-based machine learning and how to get more training data when there's not much available. It was much faster to train a model on IBM's servers than on our laptops, so we had time for a lot more iterations. In addition, we learned a lot about the malaria, such as its symptoms and its impact on African society.

What's next for Bite D.tech

Currently, the only disease Bite D.tech supports is malaria. There wasn't enough time to research more diseases. If we had more time, it would be easy to add more diseases.

Sources

https://www.unicef.org/health/files/health_africamalaria.pdf http://www.who.int/features/factfiles/malaria/en/

Share this project:
×

Updates