I am developing an AI-powered cane-mounted device that uses computer vision and edge AI to recognize objects in real time. The system runs on a compact hardware device and provides audio alerts through an embedded speaker, allowing users to understand their environment without needing an internet connection.
The current design includes an optimized TensorFlow Lite integer-quantized model for faster edge inference, a LiDAR sensor for obstacle avoidance, custom CAD-designed hardware, and a website feature that allows users to upload personal images for retraining. This makes the AI Cane more adaptable to each user’s daily environment.
The goal of the AI Cane is to improve independent mobility by helping users detect objects, avoid obstacles, and receive clear audio guidance while walking. Through hardware and software integration, the project aims to make navigation safer, faster, and more confident for visually impaired individuals.