StereoMath

Take a photo of a 3D geometry problem and get a step-by-step solution plus a 3D figure you can rotate and zoom.

Built by team Sector C at Hack TUES X (2024), where it took 2nd place overall, and kept in development until mid-2025.

How it works

1. The app takes a photo of the problem (or picks one from the gallery).
2. Google Cloud Vision reads the text.
3. The Django server asks GPT-4 for two things: a worked solution, and the figure as structured data (vertices with x, y, z coordinates and the edges between them).
4. The server validates that structure, and the app renders the figure in 3D with Three.js next to the solution.

Stack

Part	Tech
App	React Native, Expo, TypeScript, Three.js via @react-three/fiber and expo-gl, expo-camera
Server	Django, Django REST Framework, PostgreSQL, OpenAI API, Google Cloud Vision
Deployment	Heroku (server), EAS (app builds)

Run locally

# server
pip install -r requirements.txt
# create .env with DJANGO_SECRET_KEY, DATABASE_URL, OPENAI_API_KEY and EXPO_PUBLIC_GOOGLE_API_KEY (Cloud Vision)
python manage.py migrate
python manage.py runserver

# app
cd client
npm install
# set EXPO_PUBLIC_API_URL to the server's address
npx expo start

There is also a Docker setup; see procheti me.txt for the commands.

Team

	Role
Svetoslav Iliev	React Native app and the 3D figure renderer
Radoslav Dimitrov
dr-programmer
Antonio Gotev
Damyan Myashkov