ClearMyFridge | Devpost

Initial Loading Screen
First time setup screen
Log in page
Sign Up page
Preferences/ Dietary Restriction Page
Main screen, List of foods in fridge with predicted expiry date
Ingredient Selected
Ingredients/ Recipe Screen
Adding new items to the fridge
error screen

Inspiration

The inspiration for ClearMyFridge came from the universal problem of food waste. Many of us have experienced opening the fridge to find forgotten ingredients on the verge of expiring. We wanted to create a solution that reduces waste and empowers people to make the most of what they already have, turning everyday leftovers into delicious meals. This project combines our passion for sustainability, technology, and good food :)

What it does

ClearMyFridge uses LLM to generate personalized recipes based on the ingredients, their quantity, and their expiry dates that the user provides. Input the items in your fridge, and the platform provides creative and easy-to-follow recipes prioritizing ingredients nearing expiry dates. It also offers smart fridge management by tracking expiry dates and sending reminders to use your ingredients before they go bad.

How we built it

We built ClearMyFridge using:

Backend: Built REST API with Flask to handle request logic, ORM, and JSON processing, managing interactions with PostgreSQL. Database: stores uploaded ingredient data, images, user profile, and recipes. LLM Model: fine-tuned GPT-4o model for recipe generation, trained on a custom-built validation data pairs. Frontend: Unity to support mobile app integration. Hosting and Integration: Render was used to deploy the backend server and PostgreSQL, connecting CI/CD auto-deploy pipeline.

Challenges we ran into

Data Quality: Creating a high-quality dataset for training the recipe generation model was time-intensive and required careful curation. AI Fine-Tuning: Balancing creativity and practicality in recipe generation so that the results were innovative but still achievable. Time Constraints: As with any hackathon, balancing scope and execution within the limited timeframe was a constant challenge.

Accomplishments that we're proud of

Successfully fine-tuning an AI model to generate meaningful, creative recipes from a diverse range of inputs. Creating a tool that aligns with sustainability goals by helping users reduce food waste. Building a seamless integration between the front end and back end for real-time user interactions. Successfully and safely predicting the expiry date of purchased foods, allowing easy track of expiry dates for users.

What we learned

Fine-Tuning OpenAI Models: Gained experience curating datasets and fine-tuning models for personalized recipe generation. Mobile App Development: Prototyped a mobile app using Unity for seamless user interaction. Designing with Figma: Improved UI/UX skills by creating app mockups and wireframes. Hosting with Render and Flask: Learned to deploy and manage our backend efficiently using Render.

What's next for ClearMyFridge

App Release: Launch the full mobile app for widespread use. Recipe Scanning: Implement OCR to scan handwritten or printed recipes, reducing manual input. Dashboard with Social Features: Add an Instagram-like dashboard where users can share photos of their meals along with the ClearMyFridge-generated recipe, allowing others to view, try, and interact