touchGully() rather than touchGrass()

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

What I Built

Why need to touchGrass if you can just walk around your neighbourhood in you gullies(streets)?

I built GullyScout, a small AI-powered project that encourages people to step outside and explore the world around them.

The idea is simple: instead of spending all your time looking at a screen, use it to discover interesting things in your neighbourhood. This could be a street, an old building, a local shrine, or something you usually walk past without noticing.

GullyScout turns an ordinary walk into a little adventure. It's meant for anyone who wants to explore their surroundings in a more fun way.

Code

GitHub repository: (https://github.com/AbhineetB/gullyscout)

The project is still a work in progress, especially the AI model integration.

How I Built It

I built GullyScout as a web app, keeping it lightweight and focused on the main experience.

For the AI part, I'm experimenting with MobileCLIP-S0, a lightweight open-weight vision-language model. The goal is to run image analysis directly in the browser using ONNX Runtime Web and Transformers.js, rather than sending every image to a cloud API.

This approach could make the app more private, reduce running costs, and allow it to work with less dependence on an internet connection.

Getting the model to run reliably in the browser has been one of the main challenges so far, so I'm still working on that part.

Why Does Open Innovation Matter?

I wanted to experiment with AI without depending on a paid API for every request.

Open models and tools gave me the freedom to try different approaches, test things locally, and understand what happens behind the scenes. Running a model in the browser is also an interesting challenge because it comes with real limits on model size, memory, and performance.

For a project like GullyScout, open innovation makes it possible to experiment with AI in a way that could be more accessible and affordable.

My Agent Session

I used AI coding assistance while building and testing the project. I'll add my DevRelay session here if I submit one.