Fig. 1. Structure of YOLOv8 model

Fig. 2. Structure diagram of ODConv

Fig. 3. Structure diagrams of different feature fusion networks. (a) Structure of FPN; (b) structure of PAN; (c) structure of BiFPN; (d) structure of BiFPN in YOLOv8s-BiOG

Fig. 4. Structure of GAM

Fig. 5. Channel attention submodule

Fig. 6. Spatial attention submodule

Fig. 7. Structure of YOLOv8s-BiOG model

Fig. 8. Re-annotation of partial images from PIDray dataset

Fig. 9. Statistics on the number of prohibited items in SI2Pxray dataset

Fig. 10. Random X-ray images of SI2Pxray dataset and OPIXray dataset. (a)‒(c) SIXray dataset; (d)‒(f) PIDray dataset; (g)‒(i) OPIXray dataset

Fig. 11. DIoU definition diagram

Fig. 12. Confusion matrix corresponding to 5 prohibited items in SI2Pxray dataset. (a) Confusion matrix of YOLOv8s; (b) confusion matrix of YOLOv8s-BiOG

Fig. 13. Comparison of training process. (a) Precison comparison; (b) recall comparison; (c) mAP comparison

Fig. 14. Comparison images of detection results. (a)‒(e) Detection results of YOLOv8s; (f)‒(j) detection results of YOLOv8s-BiOG