I Built an AI That Tells You to Put Your Phone Down | WildWalk AI

YASH LAPTOP — Wed, 07 Oct 2026 19:34:59 +0000

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

What I Built

I built WildWalk AI, a browser-based AI outdoor exploration companion designed to encourage people to spend less time on screens and more time exploring the real world.

The idea is simple:

Use AI to get people away from their screens, instead of keeping them on one.

WildWalk AI lets users:

📷 Take or upload a photo of something they discover outdoors
🤖 Use AI to identify objects in the image
🚶 Start an outdoor walking session
🎯 Complete simple outdoor challenges
📍 Use browser geolocation
⭐ Earn points for completing challenges and making discoveries
📖 Build a local discovery history
🌙 Use a responsive interface with dark mode

The project is aimed at students, walkers, hikers, nature enthusiasts, and anyone who wants a more engaging reason to get outside.

The screen is intentionally designed to be a small part of the experience:

Start the walk
      ↓
Go outside
      ↓
Discover something
      ↓
Use AI briefly
      ↓
Put the phone away
      ↓
Keep exploring

Demo

🌐 Live Website:

https://mrhr35.github.io/WildWalk-AI/

The live demo lets you start a walk, enable location, upload a photo, run the AI analysis, complete challenges, and track your outdoor score.

Code

💻 GitHub Repository:

https://github.com/MrHr35/WildWalk-AI

The complete project is built using:

HTML
CSS
JavaScript

There is no custom backend.

How I Built It

WildWalk AI is a client-side AI application.

The project uses TensorFlow.js and the MobileNet open-source model to perform image classification directly in the browser.

The basic flow is:

User takes/uploads photo
          ↓
      JavaScript
          ↓
     TensorFlow.js
          ↓
       MobileNet
          ↓
   Image classification
          ↓
  Result + confidence
          ↓
     Discovery log
          ↓
       Points

The project also uses native browser technologies:

Geolocation API for location
LocalStorage API for saving discoveries and points
File API for image uploads
Responsive HTML/CSS for the interface

The first version intentionally avoids a backend and external AI API. The AI inference happens on the user's device/browser.

Why Does Open Innovation Matter?

Open innovation matters because the goal of WildWalk AI is not just to demonstrate an AI API.

I wanted to build something where the AI itself can eventually be modified, replaced, specialized, and adapted for the environment where the user is exploring.

A closed API could provide image recognition quickly, but an open model gives the project much more room to evolve.

For example, the current general-purpose model could eventually be replaced with an open model specifically trained on:

🇮🇳 Indian birds
🌱 Indian plants
🦋 Local insects
🌳 Regional trees
🐾 Local wildlife

That would allow WildWalk AI to become much more useful for specific ecosystems rather than depending on a generic closed AI service.

The browser-first architecture also means the project doesn't need to send every image to a custom server just to get an AI prediction.

More importantly, open AI makes experimentation possible.

The model can be changed.

The dataset can be changed.

The classification categories can be changed.

The entire AI layer can evolve with the project.

That flexibility is especially valuable for an open-source project.

My Agent Session

Optional — agent session can be added here if available.

Prize Categories

🌿 Touch Grass

WildWalk AI is primarily designed for the Touch Grass category.

The entire concept is based on using AI to encourage people to leave the screen and interact with the physical world.

Instead of:

"Ask AI another question."

WildWalk asks:

"What can you discover outside?"

Open Source AI

WildWalk AI also focuses on the open-source/open-weight AI aspect of the challenge.

The project uses a browser-based open AI model rather than relying on a proprietary AI API for its core image-classification functionality.

What's Next?

This is only the first version.

The biggest improvement I want to make is replacing the general-purpose MobileNet model with a specialized open wildlife model capable of identifying local Indian species.

Future versions could also include:

🐦 Bird-call recognition
🗺️ AI-generated nature trails
🌳 Local plant identification
🦋 Insect identification
🏆 Community challenges
📍 Location-aware species suggestions
📵 Reduced screen interaction during walks
🔌 Better offline model support

The long-term goal is to make the phone the start and end point of the experience — not the experience itself.

Final Thought

Most technology is designed to keep us looking at a screen.

I wanted to build something that does the opposite.

WildWalk AI uses AI as a bridge between the digital world and the physical world.

🌿 Go outside. Discover more.

DEV Community: YASH LAPTOP