Prohibited Object Detection using YOLOv7 on X-Ray Images of Airport Luggages (OPIXray)

Ensuring airport security and reducing risks for passengers safe travels is crucial. Conventional security inspection methods involve manual inspection of passengers luggage using X-ray scanners to identify any prohibited items. However, this manual process is prone to human error and can be time-consuming. To address these challenges, there is a need for automated security inspection systems that can enhance efficiency and accuracy. Deep learning techniques, particularly in the field of computer vision, offer potential solutions for autonomous security inspection. One such model is YOLO (You Only Look Once), a real-time object detection system capable of identifying multiple objects in images or video streams. This makes YOLO an ideal choice for airport security applications that require both speed and accuracy. This paper proposes a method for the automatic detection of prohibited objects in X-ray images, aiming to identify potential threats such as weapons, explosives, and other prohibited items in real-time. The proposed system was evaluated by conducting extensive tests on the publicly available benchmark dataset known as OPIXray. This dataset consists of labeled X-ray images that contain prohibited objects. The key component of our method involves the utilization of the YOLOv7 network, which was trained on a subset of the OPIXray dataset. The study demonstrates the effectiveness of this approach in detecting and recognizing prohibited items in OPIXray image datasets. The proposed methodology achieved a precision of 0.73, recall of 0.76, mAP@0.5 of 0.491, and mAP@0.5:0.9 of 0.397 on the OPIXray dataset. The experiments demonstrate the model's high accuracy and speed, thereby indicating its suitability for real-world airport security applications. Comparisons with other methods show that our model outperforms in terms of accuracy and speed. This paper presents a high-accuracy method for real-time airport security applications utilizing the YOLOv7 network.