Object Detection System using YOLO Model for Blind People

Real-time object detection systems offer an innovative perspective that empowers visually impaired individuals to sense their surroundings. Traditional mobility aids, such as white canes, can limit the spatial perceptions of visually impaired people, and to address this issue, intelligent systems based on Deep Learning (DL) models have been proposed. These systems are camera-based and utilize live video that is pre-processed with a Gaussian Filter and Watershed Segmentation to extract actionable visual features. Additionally, interprets an image as a topographic landscape and segments it into catchment basins based on pixel values. Object detection models, such as You Only Look Once (YOLO), are favoured for their speed and accuracy. YOLO approaches object detection as a regression problem by learning class scores and box coordinates directly. The performance metric of 92% accuracy is achieved, and to underscore the significant contributions of DL to innovative assistive technologies that enhance mobility and safety, technical advancements are made.