LostAndFound

LostAndFound

Inspiration

Before coming to the hackathon, we were still brainstorming ideas when something caught our attention—RIT’s Lost & Found space. It was disorganized, unattended, and lacked a proper system. If someone lost an item, they had to visit the space daily, hoping to find it. That’s when we thought—why not build something that simplifies this process?

What it does

Users can report lost or found items by providing details like item name, category, color, location, and description.
The system automatically matches lost items with found ones using natural language processing (NLP).
If a match is found, both the owner and finder receive an email notification.
Keeps a record of all reports, ensuring transparency and preventing false claims.

How we built it

We developed a web application where users can report lost or found items. The backend, built using Flask, handles item submissions, stores data in a Neon database, and performs natural language processing (NLP) to match descriptions. We used the SentenceTransformer ('all-MiniLM-L6-v2') model to compute similarity scores between descriptions and identify the best matches. The frontend, built with React, provides an intuitive user experience for submitting and checking item statuses.

Challenges we ran into

Faced CORS issues due to different ports; learned to configure cross-origin requests.
Switched from MongoDB to NeonDB for better support of join operations.
Struggled with styling but used Gemini API to debug efficiently.
Optimized NLP model to work with limited user input for better item matches.
Powering through exhaustion and debugging code at 4 AM was a different kind of challenge!

Accomplishments that we're proud of

Successfully built an end-to-end AI-powered matching system.
Reduced manual work by automating the lost and found process.
Created a secure and transparent platform to prevent false claims.
Implemented a scalable database and optimized search functionality.
Built a user-friendly UI that makes reporting and checking items easy.

What we learned

Implementing text similarity for matching descriptions
Integrating a machine learning model into a full-stack application
Optimizing database queries for efficient search
Building a user-friendly and secure web interface

What's next for LostAndFound

Enhancing the matching algorithm by incorporating image recognition for better accuracy.
Adding a QR code system for physical lost & found locations at RIT.
Introducing a mobile app for better accessibility and ease of use.
Improving security measures to verify users before claiming items.