Smart Glasses using Deep Learning and Stereo Camera

Blind people generally use cane or highly trained guide dogs for safe walking. In the case of the existing blind cane, there is a disadvantage that it cannot cope quickly with each changing situation. There is a cost for managing guide dogs. In this paper, we propose smart glasses using deep learning and stereo camera to overcome these disadvantages. A stereo camera calculates the distance between the user and the obstacle and then informs the dangerous level by the vibration motor and sound. In addition, the location of user is informed using LED to prevent accidents at night. We use YOLO v3 (You Only Look Once version 3), one of the deep learning algorithms, to inform the type of obstacle to the user. The experimental results indicate that the proposed smart glasses overcome the existing blind guidance system disadvantages and may be helpful when the user recognize and avoid obstacles.