SightMate: A Vision-Based Assistive Navigation System for the Visually Impaired

Visual impairment significantly restricts independent mobility and environmental awareness for millions of individuals worldwide, creating challenges in safely navigating unfamiliar environments. Traditional assistive tools such as white canes and guide dogs provide limited perception of distant obstacles and complex surroundings. Recent advances in computer vision and mobile artificial intelligence have enabled the development of intelligent assistive technologies capable of interpreting environmental scenes in real time. This paper presents SightMate, an AI-driven vision-based assistive navigation system designed to support visually impaired users through real-time environmental perception and auditory guidance. The proposed system integrates multiple perception and interaction layers, including camera-based scene acquisition, obstacle detection using the YOLOv8 deep learning object detection framework, and walkable path segmentation using the Fast-SCNN semantic segmentation network to identify safe navigation regions. A zone-based navigation decision module analyzes spatial obstacle distribution across left, center, and right regions of the captured frame to generate intuitive movement instructions. To improve robustness in dynamic environments, motion filtering and frame stability analysis are incorporated to reduce false detections caused by camera movement. The system also includes accessibility features such as voice-command interaction, Optical Character Recognition (OCR) for environmental text reading using ML Kit, and a six-dot Braille keyboard interface that enables tactile user input. The complete solution is implemented as a Flutter-based mobile application utilizing TensorFlow Lite for on-device inference, enabling low-latency real-time processing without dependence on continuous internet connectivity. Experimental evaluation demonstrates reliable obstacle detection and walkable path identification under diverse environmental conditions while maintaining computational efficiency suitable for smartphone deployment. The proposed framework bridges the gap between mobile computer vision research and practical assistive technology by delivering a scalable, real-time, and user-centric navigation assistance platform for visually impaired individuals. Keywords— Assistive Navigation, Visual Impairment, YOLOv8, Fast-SCNN, Object Detection, Semantic Segmentation, OCR, Braille Interface, Voice Assistance, TensorFlow Lite, Mobile AI, Computer Vision.