INSIGHT: In-device Navigation and Scene Interpretation Glasses for Human-centered Travel of Low-vision Users
Published in EECS 473 Advanced Embedded Systems – University of Michigan, 2025
Group Members: Haobo Fang, Guanyu Xu, Ruopu Dong, Zhuoyang Chen, Jinlin Li, Yizhe Shen.
Project Status: Completed. For more implementation details, please refer to our final report and the github repo1 (Smart Glasses) or repo2 (Jetson Nano).
Abstract: INSIGHT is a low-cost smart assistive eyewear system designed to improve safe mobility for people with visual impairments. Built from commodity hardware and compact on-device models, the system can be mass-produced for around $300—significantly below many existing commercial solutions. Using wireless communication between the glasses and a companion processing unit, INSIGHT provides GPS-guided navigation, obstacle-aware guidance, and natural scene descriptions in real time, all without relying on cloud connectivity. This fully local processing pipeline ensures privacy, low latency, and robust offline performance, illustrating how embedded intelligence in a practical wearable form factor can enhance independence and everyday quality of life for low-vision users.