NavAI - Reimagining Navigation Through AR

🏆 Winner of "Best Use of Gemini AI" at ImmerseGT 2024 - Georgia Tech's Premier VR/XR Hackathon

An AR-powered navigation platform that transforms urban exploration through natural voice commands, dynamic AR mapping, and AI-generated insights. Using Snap Spectacles and Gemini AI, NavAI delivers a hands-free, immersive wayfinding experience for modern explorers.

---

Table of Contents

• Overview
• Tech Stack
• Directory Structure
• Modules and Functions
  • Voice Interaction & Gemini AI
  • Dynamic AR Mapping
  • Snapshot Analysis
• Usage
• License

---

Overview

NavAI reimagines navigation by integrating immersive AR with AI-driven natural language understanding and image analysis. With voice-activated queries and real-time AR map overlays, users can effortlessly discover points of interest and obtain contextual information—all without ever pulling out their phone.

---

Tech Stack

Hardware

Snap Spectacles — AR-enabled smart glasses
Built-in Camera & Mic — For seamless image capture and voice input
Location Sensors — Precise geolocation for dynamic mapping

Core Technologies

TypeScript — Strongly-typed development environment
Snap AR SDK — AR content rendering and spatial tracking
SIK (Spectacles Interaction Kit) — Gesture recognition and interaction

AI & APIs

Google Gemini AI — Natural language processing and visual analysis
Google Places API — Dynamic retrieval of nearby points of interest
Custom Flask Backend — Image processing and AI integration

---

Directory Structure

/Outdoor Navigation
├── Assets/                      # Main project assets and code
│   ├── VoiceController.ts      # Voice command processing
│   ├── SpeechToText.ts         # Speech recognition integration
│   ├── Scripts/                # Core TypeScript/JavaScript files
│   │   ├── geminiHandler.ts    # Gemini AI integration
│   │   ├── PlacesWrapper.ts    # Places API wrapper
│   │   ├── MapUIController.ts  # Map interface controls
│   │   ├── MapManipulation.ts  # Map interaction handling
│   │   ├── MapConstants.ts     # Configuration constants
│   │   ├── NorthIndicator.ts   # Compass functionality
│   │   ├── PinchController.ts  # Gesture recognition
│   │   ├── PopupController.ts  # Information display
│   │   ├── QuestMarkController.ts  # Navigation markers
│   │   ├── UICollisionDetector.ts  # UI interaction handling
│   │   ├── MapMessageController.ts  # Map messaging system
│   │   ├── MapContainerController.ts # Map container management
│   │   └── Logging.ts          # Logging utilities
│   ├── MapComponent/           # Map visualization system
│   │   ├── Scripts/           # Map-specific scripts
│   │   ├── Prefabs/           # Reusable map elements
│   │   ├── Materials/         # Map styling assets
│   │   └── Textures/          # Map visual resources
│   ├── SpectaclesInteractionKit/  # SIK integration
│   ├── Device Camera Texture   # Camera input handling
│   ├── VoiceInput             # Voice input configuration
│   └── SoundOutput            # Audio output handling
├── Support/                    # Project support files
├── Workspaces/                # Lens Studio workspace configs
├── README-ref/                # Documentation assets
├── tsconfig.json              # TypeScript configuration
└── OutdoorNavigation.esproj   # Project configuration file

Key Components

• Voice Processing: VoiceController.ts and SpeechToText.ts handle wake word detection ("Hey Alice") and voice commands
• Map System: Multiple controllers (MapUIController, MapManipulation, etc.) manage the AR map experience
• AI Integration: geminiHandler.ts coordinates between Gemini AI and the system, while PlacesWrapper.ts handles Places API interactions
• Interaction: PinchController.ts and PopupController.ts manage user input and feedback
• Supporting Systems: Includes camera, voice input, and sound output configurations

The project follows a modular architecture, separating concerns between voice processing, map visualization, AI integration, and user interaction. This structure enables easy maintenance and future enhancements of individual components.

---

Modules and Functions

Voice Interaction & Gemini AI

• Purpose: Interprets natural language queries (e.g., "Find me a quiet coffee shop") via voice commands.
• Functionality: Uses Gemini AI to map user intent to relevant location categories, generating structured requests for the Places API.

Dynamic AR Mapping

• Purpose: Renders a real-time, interactive AR map overlay displaying nearby points of interest.
• Functionality: Updates map pins dynamically as the user moves, ensuring location-aware navigation.

Snapshot Analysis

• Purpose: Captures and processes visual data using simple gesture controls.
• Functionality: Uses a pinch gesture to capture surroundings and leverages OCR and AI to generate rich, contextual descriptions of buildings or landmarks.

---

Usage

1. Activate Voice Command: Say the trigger phrase (e.g., "Hey Alice") and ask for your desired location.
2. Explore AR Map: Watch as dynamic location pins appear over your real-world view.
3. Capture Insights: Use a pinch gesture to capture a scene and receive immediate AI-generated insights.

---

License

This project is licensed under the MIT License.