VisuBot: A Real-Time Object Describer for Visually Impaired Users Using YOLOv8

Visual impairment significantly restricts independent mobility, environmental awareness, and social participation, particularly in developing countries such as India where access to assistive technologies remains limited. Traditional navigation aids like white canes provide only basic obstacle detection and lack contextual intelligence. Recent advances in artificial intelligence, deep learning, and computer vision have enabled the development of intelligent assistive systems capable of interpreting visual scenes in real time. This paper presents VisuBot, a real-time object describer designed specifically for visually impaired users. The proposed system employs the YOLOv8 deep learning model trained on the COCO dataset to detect navigation-relevant objects in real time. Detected objects are analyzed to estimate direction using a clock-based spatial mapping technique and distance using bounding box area approximation. The processed information is converted into natural language audio feedback using an offline text-to- speech engine. VisuBot operates entirely offline on low-cost edge devices, ensuring affordability, portability, and reliability in rural and low-resource environments. Experimental observations indicate that VisuBot significantly enhances situational awareness, user confidence, and navigation safety. The system demonstrates low-latency performance with response times under 200 milliseconds, making it suitable for real-world deployment in diverse environmental conditions.