IoT Smart Glasses with TinyML & Edge AI
Real-time assistance for visually impaired individuals in Sri Lanka through facial recognition, Sinhala voice navigation, and text reading—all powered by affordable, offline-capable technology.
System Overview
Complete architecture and workflow demonstration
Three Integrated Components
A comprehensive assistive system combining facial recognition, voice navigation, and text reading—all designed specifically for the Sri Lankan visually impaired community
Real-Time Facial Recognition
Identify known individuals with 92% accuracy. Dynamic enrollment via voice commands. Whisper feedback for privacy.
Technologies:
FaceNet, MobileFaceNet, TensorFlow Lite
Sinhala Voice Navigation
Speech-to-text commands (75-85% accuracy) with ultrasonic obstacle detection up to 1m. Real-time audio warnings.
Technologies:
Whisper STT, HC-SR04 sensors, Android TTS
Sinhala Text Recognition
OCR for printed Sinhala text with document classification (exam papers, newspapers, forms, notes, stories, words).
Technologies:
Google ML Kit, Tesseract, CNN classification
System Architecture
Edge AI processing on Raspberry Pi 5 for privacy, affordability, and offline functionality
Mobile Application
Android/React Native app with voice commands, frame filtering, and Bluetooth communication
- Voice feedback system
- Accessibility features
- Sinhala & English support
Edge Server (Raspberry Pi 5)
Flask REST API running as systemd service for local AI processing
- Face recognition
- Obstacle detection
- OCR processing
Smart Glasses Hardware
Pi Camera Module 2 (8MP) mounted on lightweight frame with portable power
- Real-time capture
- Low latency
- Wearable design
Hardware Components
Affordable, accessible hardware totaling ~Rs. 70,000 (compared to OrCam's $3,500+)
Raspberry Pi 5
- Broadcom BCM2712
- 8GB RAM, 8 TOPS
- 2.4GHz Quad-core
~Rs. 43,000
Pi Camera Module 2
- 8MP Sony IMX219
- 1080p video
- Wide-angle lens
~Rs. 6,500
HC-SR04 Ultrasonic Sensors
- 2cm-1m range
- Real-time detection
- Multiple sensors
~Rs. 500 each
Power & Connectivity
- 27W USB-C supply
- Bluetooth/Wi-Fi
- Portable battery
~Rs. 10,000
Why Our Solution Stands Out
Designed specifically for the Sri Lankan context with privacy, affordability, and accessibility in mind
Privacy First
100% on-device processing. No cloud dependency. Your data stays with you.
Offline Capable
Works without internet. Perfect for rural areas with limited connectivity.
Sinhala Support
Native Sinhala speech recognition, TTS, and OCR—first of its kind.
Affordable
~Rs. 70,000 total cost—10x cheaper than commercial alternatives like OrCam.
Research Impact
This research advances intelligent, real-time assistance systems by demonstrating that IoT Smart Glasses can operate efficiently in low-power, privacy-sensitive environments while maintaining high accuracy and reliability for visually impaired individuals in developing regions.
Empowering Independence Through Technology
Explore our comprehensive research project and discover how edge AI and TinyML are revolutionizing assistive technology for visually impaired individuals in Sri Lanka.