Vision-Link AI Hub：打造供視障人士使用的離線邊緣 AI 智能眼鏡

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine

What I Built

Vision-Link AI Hub is an assistive smartglasses solution powered by 100% offline edge-AI. Utilizing custom YOLO and MobileNet models trained via Roboflow and deployed via TensorFlow Lite on Android/UVC hardware, it detects real-time environmental obstacles and translates them into instant spoken audio warnings for visually impaired users—requiring zero cloud latency or internet connectivity.

Demo

https://github.com/Vision-Link-AI-Hub

Code

Check out our core repository and mobile implementation on GitHub: https://github.com/Vision-Link-AI-Hub

How I Built It

We built Vision-Link AI Hub using open-weight models (SSD MobileNet / YOLO), TensorFlow Lite for on-device Android inference, and UVC camera integration. By keeping inference completely on the edge, the device operates seamlessly in airplane mode, ensuring user privacy and instant response times under 100ms.

Why Does Open Innovation Matter?

Open innovation and open-weight models allow developers to deploy life-changing assistive technologies directly to local hardware without depending on costly cloud APIs or stable internet connections. For visually impaired individuals in developing regions, offline capability is not just a feature—it is a necessity for safety and independence.

My Agent Session

Developed with the assistance of advanced AI coding workflows, optimizing our TFLite Android integration and edge-AI pipeline.

The Spirit of It

Building Vision-Link AI Hub was driven by a deep passion to give sight through sound. Knowing that a visually impaired person can walk independently, avoid obstacles safely, and hear instant real-time audio warnings without needing an internet connection brings life-changing impact to reality.

Prize Categories

Open Source AI, Healthcare & Assistive Technology, Edge Computing.