CFIPC-YOLO for Concealed Object Detection in Millimetre Wave Images

Human body security inspection is a critical task, and most existing methods rely on millimetre-wave images to detect concealed objects. However, detecting small and dim objects remains a significant challenge due to their subtle visual characteristics. To address this issue, this study proposes an efficient model, CFIPC-YOLO (contextual features integrated progressive convergence network based on the YOLOv8), which integrates local and contextual features into a progressively converging framework based on YOLO. Experimental results show that CFIPC-YOLO achieves a 5.3% improvement in AP50 compared to the YOLOv8 baseline, while simultaneously reducing the number of parameters by 17%, confirming the model's superior performance.