On the Use of Deep Learning for the Detection of Firearms in X-ray Baggage Security Imagery

X-ray imagery security screening is essential to maintaining transport security against a varying profile of prohibited items. Particular interest lies in the automatic detection and classification of prohibited items such as firearms and firearm components within complex and cluttered X-ray security imagery. We address this problem by exploring various end-to-end object detection Convolutional Neural Network (CNN) architectures. We evaluate several leading variants spanning the Faster R-CNN, Mask R-CNN, and RetinaNet architectures. Overall, we achieve maximal detection performance using a Faster R-CNN architecture with a ResNet
₁₀₁
classification network, obtaining 0.91 and 0.88 of mean Average Precision (mAP) for a two-class problem from varying X-ray imaging dataset. Our results offer very low false positive (FP) complimented by a high accuracy (A) (FP=0.00%, A=99.96%). This result illustrates the applicability and superiority of such integrated region based detection models within this X-ray security imagery context.