NAVISIGHT: A Deep Learning and Voice-Assisted System for Intelligent Indoor Navigation of the Visually Impaired
Avula Sachin, Aryan Penukonda, M. Naveen, Pramod Gurunath Chitrapur, Praveen Kulkarni, B M Chandrakala
Abstract
Indoor navigation is still a significant issue for the blind, restricting their mobility and independence. White canes and guide dogs are valuable aids but cannot communicate advanced spatial information. Over the last decade, advances in computer vision and AI have enabled more advanced navigation technology. This paper introduces an AI -based room and object recognition system that can facilitate the indoor navigation of blind and partially sighted people. Our method employs deep learning models, in this case, YOLO for object recognition, to categorize room types by identified objects. This research also have depth detection for estimating objects' distances from the user and voice output in real-time, allowing them to navigate better. Voice narration is also incorporated into the system for giving verbal descriptions, alerting users to obstacles and routes around them. This paper also present a well-organized dataset that is specifically designed for indoor scene classification with the best classification accuracy. The solution aims to fill the gap between current navigation assistance and complete autonomous support systems, providing an easy to use and practical device for visually impaired people.