Enhancing Transportation Security: Automated Prohibited Object Detection for Baggage Inspection Leveraging YOLO-NAS

The substantial increase in demand for travel has resulted in significant congestion and prolonged wait times across various transportation sectors, prominently in the aviation industry. The primary cause of these delays can be attributed to the time-consuming manual inspection of baggage to detect prohibited objects. This predicament underscores the imperative need for the deployment of a sophisticated and efficient automated system to comprehensively identify prohibited objects in real-time. In response to this, we have developed a one-stage detection system using You Only Look Once – Neural Architecture Search (YOLO-NAS), a cutting-edge iteration within the YOLO family. An extensive experimentation was conducted on the SIXray dataset, which yielded remarkable efficacy, boasting a mean Average Precision (mAP) score of 0.924. Our research findings reveal the potential of YOLO-NAS to significantly improve the accuracy and speed of prohibited object detection in baggage, thereby mitigating delays and congestion within transportation hubs. Furthermore, our research extends to comparative analysis, pitting against its predecessor, YOLOv8, where YOLO-NAS demonstrated its superiority by surpassing YOLOv8’s performance by 12%. This utilization of YOLO-NAS underscores the potential for transformative improvements in baggage inspection and serves as a testament to the continued advancement of automated detection systems within the domain of transportation security.