EchoGuide: AI-Powered Audio Guide for the Visually Impaired

Overview

This project enables real-time scene understanding and spoken navigation for visually impaired users using computer vision and AI. Currently still under development.

Pipeline Architecture

Camera Feed
↓
YOLOv8 Object Detection
↓
Detected Objects + Bounding Boxes
↓
Spatial Analysis (position, distance, urgency)
↓
LLM Narrator (Llama 3.1 – Groq API)
↓
Navigation Instruction (text)
↓
Text-to-Speech (gTTS)
↓
Audio Output
↓
Flask Web App (User Interface)

Demo

Coming soon!

Tech Stack

Backend

• Python
• Flask
• OpenCV
• YOLOv8 (Ultralytics)
• Groq API
• gTTS

Frontend

• HTML/CSS/JS
• AJAX for fetching narration + audio
• Live MJPEG video stream

AI Models

• YOLOv8n — real-time object detection
• Llama 3.1 / LLM — intelligent navigation guidance
• gTTS — text-to-speech output

Installation

1) Clone the repo:
git clone https://github.com/TibyanKhalid/EchoGuide.git
cd EchoGuide

2) Create virtual environment:
python -m venv venv
venv\Scripts\activate

3) Install required libraries and dependencies:
pip install -r requirements.txt

4) Set a Grok token:
setx GROQ_API_KEY "your_token_here"

5) Run the webapp:
python app.py

Limitations

• Requires stable internet (LLM API)
• Currently supports English Language only
• Works best in well-lit environments
• Navigation accuracy depends on YOLO bounding boxes