Food-App

Inspiration

Growing up, many of us were taught the basic food groups and the food pyramid, but as adults, we rarely think about balanced nutrition in a simple, visual way. Most health apps today focus on calorie counting, strict dieting, or complicated tracking. We wanted to solve this problem by bringing back the simplicity we had as kids, just knowing our food groups and eating a balanced plate. Our app helps people understand what they’re missing. Whether it’s fruits, vegetables, grains, protein, or dairy, and gives friendly guidance in real time so staying healthy feels natural, not stressful.

What it does

Our app uses AI to: Identify foods from images and give helpful nutritional guidance Track progress through streaks (to be implemented), and positive habit-building Provide personalized suggestions based on what the user eats each day, with a chat bot

How we built it

We used NextJs for the front end and fast API for backend, connected with pytorch ML functionality and Supabase for the database.

Challenges we ran into

Challenges we ran into One of the biggest challenges we ran into was training the pytorch ML model properly with a diverse group of datasets and training it properly with good weights. Another big problem was managing all the containers and learning docker in such a time sensitive and fast environment. Furthermore, connecting all the endpoints together and managing merge conflicts across everyone was tough but we pushed through with everything.

Accomplishments that we're proud of

Building an end-to-end working app in a short time A machine learning model that detects what nutrients the food has Connecting all the API endpoints to the frontend

What we learned

With all the separate frontend, backend, and ML components, it was hard bringing them all together to be ran with one command, instead of starting the frontend, backend, and ML servers separately. We learned that Docker is a very helpful service for tackling problems like this.

What's next for Food-App

There are many features we had in mind that we didn't get to implement, such as user authentication, user-to-user connection, streak tracking, and some other small database optimizations. [tbd]