MixChannel_YOLO: Traffic sign detection based on mix channel attention yolov8
Yu Zhou, Chuanhao Wei, Hongmei Liu, Dezhao Kong · 2024
Traffic sign recognition is dedicated to ensuring the safety of autonomous driving. Inspired by YOLOv8, this paper proposes a new model to address the difficulty in balancing accuracy and efficiency that exists in current algorithms for traffic sign recognition. We introduce a novel hybrid channel attention mechanism and incorporate Slim-Neck to achieve higher computational cost-effectiveness in the detector, and we employ WIOU loss to provide a wise gradient gain distribution strategy. Our network achieves an [email protected] of 54.7% on the CCTSDB2021 dataset, which represents a 3% improvement over YOLOv8.