VisionNav

Overview

VisionNav transforms smartphone camera input and Google Maps data into clear haptic guidance. Designed for people with visual/hearing impairments and for noisy or visually demanding environments, it provides reliable, safe, and accessible navigation using simple, understandable feedback.

Features

• Turn-by-turn navigation enhanced with vibration cues;
• Obstacle detection and avoidance using YOLOv8-based vision models;
• Path position recognition via a lightweight CNN classifier;
• Mobile UI for destination search, route view, and QR code sharing;

System Architecture

VisionNav consists of a mobile app, on-device computer vision, and a wearable haptic device (M5Stack). The phone camera detects obstacles and path cues locally, while Maps data provides route-level guidance. Haptic patterns encode direction and urgency.

Interaction

• Single-beat: continue forward
• Double-beat left/right: turn guidance
• Long pulse: caution/obstacle detected

Model

We use a compact YOLOv8 variant for obstacle detection and a lightweight CNN for path position classification, both optimized for real-time inference on-device.