Mobile QR Code
Title A Detection Performance Comparison Study of Threat Inspection of Passenger Baggages using Backbone Models
Authors 편집부(Editor)
DOI https://doi.org/10.5573/ieie.2025.62.2.74
Page pp.74-82
ISSN 2287-5026
Keywords Security inspection; Threat detection; X-ray
Abstract In this paper, we compared the object detection performance on an X-ray image-based dataset using a backbone model to improve the detection performance of hazardous item on passenger baggage X-ray images. We tested the detection performance by applying various types and sizes of backbone models to the DINO (DETR with Improved DeNoising Anchor Boxes) detector, which has recently shown excellent object detection performance. The InternImage-S backbone showed a detection performance of 88.4 mAP0.5 for the SIXray-10 dataset. We explored the possibility of monitoring the detection performance during training by comparing the change in detection performance during the training process and the detection performance between backbone model sizes.