VisionMate
Real-time assistive technology for visually impaired individuals.
The Problem
Visually impaired individuals in Nepal face significant barriers navigating daily environments — limited access to real-time assistive tools, poor indoor wayfinding, and near-zero on-device AI support in the local context.
What They Built
VisionMate is a Flutter mobile application that provides real-time object detection, indoor navigation assistance, and on-device scene narration for visually impaired users. The app uses TensorFlow Lite for on-device inference (no server round-trip needed), FastAPI for backend coordination, Groq API for fast natural-language scene descriptions, and Firebase for user state. The result is a responsive, offline-capable assistive tool designed for real-world use.
Impact
VisionMate won SXC Sandbox 2.0 outright. The project was maintained and expanded after the hackathon, with continued development led by the team post-event. It stands as one of the strongest demonstrations of practical AI applied to accessibility at a Nepali student hackathon.