Deep Learning Based Tool (Dblt) to Assist Movement Inside Closed Spaces for People Affected with Blindness

Abstract

Given the large population of blind and visually impaired individuals (VIIs), there is an increasing need for intelligent assistive tools that can provide real-time warnings about potential collisions. Navigation for VIIs using conventional aids like white canes and guide dogs has limitations implying the need for them to overcome movements in physical environments. VIIs have issues even in own their homes. They may find it difficult to navigate interior parts of homes including corridors or confront potential hazards. Hence, this work suggests an aid that can help VIIs walk based on their surroundings. The schema called Deep Learning Based Tool (DBLT) to help VIIs can detect obstacles that block their paths and help in assisting their movements from one place to another including homes. The schema uses generated images of obstacles and notifies the user of actions like moving forward or left or right. This work’s evaluation results show promise in implementations and if attached to smart devices that capture images of front sides, DBLT warns VIIs about obstacles, allows them to walk straight with referral actions.