Semantic Segmentation For Prohibited Items In Baggage Inspection

The X-ray screening system is crucial to protecting the safety of public spaces. However, automated detection in baggage inspection is still far from practical application. Most detection tasks rely mainly on humans. In this paper, the detection of prohibited items is regarded as a semantic segmentation task. Considering some characters of security imageries, we propose a segmentation net with novel dual attention, which could capture richer features for refining the segmentation results. Our model could not only automatically recognize the class of prohibited items but also locate prohibited items in baggage. It could facilitate the security staffs to carry out inspection. To validate the effectiveness of our proposed model, extensive experiments have been conducted on the real X-ray security imageries datasets. The experimental results show the net achieves super performance (mIoU of 0.683).