We were really aiming to learn something new at this hackathon - something that neither of us had touched before. We decided to attempt machine learning on photos of overweight individuals to try and diagnose obesity type.

What it does

The app is meant to scan an image of the user and to then diagnose their obesity type. With that knowledge, the app can then make personalized exercise/diet recommendations to optimize a user's weight loss.

How we built it

The source code for machine learning was provided by a mentor, and the app was built with react native framework and the expo camera view component

Challenges we ran into

The machine learning code provided to us by some mentors had many errors popping up - as well as that we experiment with a cloud-based kernel using Kaggle; something we'd never touched before. Additionally, no datasets for our problem existed, so we had to make our own.

Accomplishments that we're proud of

We managed to get an app up and running that can take/access photos, then make recommendations based on that photo. We also made a little progress with machine learning and learned how to make datasets.

What we learned

We now know a lot more about machine learning. Through attending (all) the workshops, talking extensively with sponsors, and heavily consulting mentors, we have a much better understanding of how machine learning, at its core, really works.

What's next for Weight-free

The biggest thing would be improving the machine-learning algorithm, which is still in very early stages Other improvements could be location trackers in order to give more personalized advice, and a separate neural network that could track the specific stress levels and eating habits of each individual, making for a more personal fitness tracker.

Share this project: