CIFE-YOLOv8: A Dense Pedestrian Detection Model Based on Improved YOLOv8

In response to the current challenges of pedestrian detection in dense scenes, we propose CIFE-YOLOv8, a method tailored for this purpose. Firstly, we conduct benchmark tests on the YOLOv8 model to demonstrate its potential in dense pedestrian detection tasks. Secondly, we introduce CIFNet into the neck to enhance the utilization of feature information and focus on small-scale pedestrians within the feature fusion network. Finally, EMA is introduced into the backbone to encode global information and further aggregate pixel-level features through dimension interaction. Experimental results on the WiderPerson dataset demonstrate a significant improvement of our proposed dense pedestrian detection method, CIFE-YOLOv8, compared to YOLOv5. Specifically, it achieves a 1.8% increase in mAP50 and a 0.9% increase in mAP<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">50-95</inf>, indicating superior performance of CIFE-YOLOv8 in dense pedestrian detection tasks.