Real-Time School Bus Safety Monitoring Using YOLO-Based AI-Assisted Computer Vision System; The Use of AI-Assisted Cameras on School Buses: Addressing Systemic Gaps in Child-Check Safety Protocols

In May 2021, a five-year-old child in Houston, Texas died after being left behind on a school bus. This heartbreaking incident, along with others like it, exposed how fragile and outdated school transportation safety systems can be. Motivated by this tragedy, I pursued a capstone project proposing a technical solution to monitor school bus interiors in real time using computer vision. In parallel, I developed an STS research paper investigating the institutional, human, and technological gaps that contribute to recurring failures. Together, the projects address the same critical issue from complementary perspectives: one through engineering innovation, the other through sociotechnical analysis.

My capstone project, Real-Time School Bus Safety Monitoring Using YOLO-Based AI-Assisted Computer Vision System, seeks to address the problem of children being accidentally left behind in school buses. Traditional safety methods like manual headcounts and motion sensors have proven unreliable. To address this, I developed an embedded system using YOLOv5, a fast object detection algorithm, on Jetson Nano edge devices. The system analyzes real-time video from onboard cameras to detect students and trigger alerts if passengers are left on the bus after the route ends.

The project demonstrated that combining infrared cameras, YOLO detection, and cloud-based alert systems can provide a scalable and cost-effective safety net. The system sends alerts to both the bus driver and administrators via a cloud platform. Testing showed it can reliably detect children even under occlusion or poor lighting. Challenges such as reflections and hardware limitations were addressed through model quantization and adaptive preprocessing. The results support real-world deployment, and future work includes pilot testing and system simplification for broader implementation.

My STS research explores the same issue from a sociotechnical perspective. The central research question asks how human factors, institutional policies, and current technologies combine to create vulnerabilities in school bus safety, and whether AI-assisted systems can address these gaps. Using Actor-Network Theory and the sociotechnical triangle, I examined the relationships among drivers, school policies, regulations, and safety technologies. My methods included policy analysis, literature review, and scenario modeling based on real-world cases like the 2018 Houston tragedy.

The study found that technical failures are often symptoms of deeper systemic issues such as high driver turnover, limited funding, inconsistent policy enforcement, and insufficient training. Districts with tighter budgets and lower maintenance spending report significantly more safety lapses. Evidence from audits and incident data shows AI systems have great potential, but only if supported by training, maintenance, and legal considerations like FERPA compliance. The research concludes that while AI can enhance safety, it must be integrated into a broader framework of operational and institutional reform to be truly effective.

School of Engineering and Applied Science

Bachelor of Science in Computer Science

Technical Advisor: Briana Morrison

STS Advisor: Pedro Francisco

Technical Team Members: Hassan Al Moalim