Adaptive Vision AI: Revolutionizing Navigation for the Visually Impaired with Real-Time Assistive Feedback

Abstract

The research work introduces a vision-based adaptive assistance system intended to aid the mobility and awareness of the blind. Real-time computer vision techniques like obstacle detection with distance estimate, staircase detection, signboard recognition, cash identification, and document text extraction with summarization are all included in the suggested system. The You Only Look Once version 8 model is used for object identification, and the Mixed Depth Scale model is used for depth estimates. Additionally, the system uses the Oriented FAST and Rotated BRIEF algorithm for cash recognition and Optical Character Recognition for textual information extraction. Context-aware help is made possible by the delivery of the derived information via haptic and audible feedback. Real-time performance is ensured by the system’s implementation on a portable Raspberry Pi-based edge computing platform. The technology improves the independence and safety of visually impaired users by achieving high accuracy and robustness in a variety of contexts, according to experimental results.



Received: 21 January 2026 | Revised: 7 May 2026 | Accepted: 28 May 2026



Conflicts of Interest

The authors declare that they have no conflicts of interest to this work.



Data Availability Statement

Data sharing is not applicable to this article as no new data were created or analyzed in this study.



Author Contribution Statement

R. Nirmalan: Conceptualization, Methodology, Validation, Investigation, Resources, Data curation, Writing – review & editing, Visualization, Supervision, Project administration. C. P. Amirdha Suba: Conceptualization, Methodology, Software, Formal analysis, Writing – original draft, Visualization. M.A.M. Atchaya Durga: Conceptualization, Methodology, Software, Formal analysis, Writing – original draft, Visualization.