An Intelligent Model for Detecting Obstacles on Sidewalks for Visually Impaired People

Blind individuals face significant challenges while navigating outdoor environments, particularly on sidewalks shared with the general public. Identifying and avoiding common objects, such as trash cans, benches, or bike racks, is crucial for their safety and independence. This paper proposes an intelligent system that employs the You Only Look Once (YOLO) object detection algorithm to detect common objects on blind sidewalks. Our work mainly proposes a new dataset (OOD) that contains 10,000 images and 29,779 annotated instances, and 22 different types of classes. We train the dataset with the state-of-the-art object detection models, YOLOv5 and YOLOv8, to accurately recognize objects such as benches, trash cans, street signs, fire hydrants, and more. We analyze and compare these models in detail and then deploy the optimal model on Raspberry Pi. By utilizing real-time video input from camera-equipped devices, the system provides auditory feedback to alert users about the presence of objects, thereby aiding navigation in outdoor environments.