A Computer Vision Model to Support Individuals with Disabilities Within University Campuses

This study introduces MCOF, a multi-camera, computer vision-based system designed to assist visually impaired individuals with mobility on university campuses. The system operates locally and achieves the following performance metrics: (i) detecting body points within 1.2. 10–2seconds, (ii) identifying people, objects, and animals in 2.4. 10–2seconds, and (iii) detecting movements in 5.1. 10⁻² seconds. These results were obtained using a GTX 1,660 GPU with up to 6 cameras (or a 6,112MB stream) running concurrently. According to the MCOF architecture, events trigger tickets that are sent to an external information system, which can then implement its own safety and personnel protocols. Additionally, MCOF includes modules to handle electrical and network failures and features an obstruction detection routine for the cameras.