IAN: Intelligent Accessibility Navigator
A real-time visual-assistance system for the visually impaired utilizing Gemini Vision capabilities, deployed on Google Cloud Run.
Gemini 2.5 VisionGoogle Cloud RunText-to-SpeechPythonREST API
Problem Statement
Visually impaired individuals face daily challenges navigating physical environments without real-time contextual information. Existing assistive tools often provide limited scene understanding. IAN leverages multimodal AI to provide natural language descriptions of visual scenes in real-time.
Architecture
Processing Pipeline
Camera InputReal-time video
Gemini Vision2.5 Flash analysis
Scene DescriptionNatural language
TTS OutputSpoken guidance
Visual Assets
Demo recording and accessibility testing screenshots coming soon.
Measurable Results
Real-Time
Scene description pipeline
Cloud Run
Serverless deployment
Multimodal
Vision + Language + Speech
Accessible
Designed for visually impaired users
What I Learned
- Designing for accessibility-first rather than retrofitting — the entire architecture serves users with visual impairments.
- Using Gemini Vision for real-time scene understanding and converting visual data to natural language.
- Deploying latency-sensitive AI services on Cloud Run for auto-scaling serverless inference.
- TTS integration challenges — balancing response speed with natural-sounding spoken output.