The X-ray Contraband Detection Dataset is a dataset released in 2025 by South China Normal University in collaboration with Hong Kong Polytechnic University and the University of Saskatchewan. It is designed for detecting contraband targets in X-ray security inspection scenarios. The related research paper is as follows: Balanced X-ray Security Dataset and Enhanced YOLO for Contraband DetectionThe design aims to improve the detection model's ability to identify complex and crowded security images, especially addressing real-world problems such as class imbalance and sample scarcity.

This dataset contains 13,728 X-ray security inspection images, covering 12 common categories of prohibited items. An approximately balanced sample distribution was constructed for each category, with approximately 1,500 images per category, effectively mitigating the class imbalance problem commonly found in traditional security inspection datasets. The dataset is built using SIXray and PIDray, and class balance is achieved through CSAF enhancement and downsampling strategies.