From Pixels to Paths: SIGHT - A Vision-Based Navigation Aid for the Visually Impaired

Visual impairment significantly challenges daily life, limiting mobility and independence. Traditional devices such as canes have provided limited detection capabilities, and the applications developed have yet to give the desired mobility and reliability. In this work, we aim to solve the problems of visually impaired individuals using image processing technologies and propose a mobile application called "SIGHT" that provides real-time object recognition and guidance. SIGHT leverages YOLOv8 for object detection and MiDaS for monocular depth estimation identifies objects by processing live camera images and provides audio feedback. Unlike GPS-based solutions, SIGHT functions in both indoor and outdoor environments. Key features include real-time object detection for accessibility and enhanced spatial awareness that requires no additional hardware on existing smartphones. The application, moreover, integrates voice interaction through Android's Text-to-speech and Speech-to-text services so users can interact with voice commands and receive auditory feedback. Apart from development, we also evaluated the performance of the proposed approach on different smartphone models with varying optimization strategies. The obtained results demonstrate the effectiveness of SIGHT in providing real-time object recognition and navigation assistance for visually impaired individuals.