山东理工大学学报（自然科学版）2026，Vol.40Issue(5)：15-21,7.
改进YOLO11的轻量化驾驶员疲劳检测算法
Improved lightweight driver fatigue detection algorithm for YOLO11
张洋 1朱泽德 2李宸翔 1王道斌1
作者信息
- 1. 武汉科技大学 汽车与交通工程学院,湖北 武汉 430065
- 2. 中国科学院 合肥物质科学研究院,安徽 合肥 230031
- 折叠
摘要
Abstract
To address the challenges posed by large number of parameters and high computational re-quirements of driver fatigue detection models,which make deployment on mobile devices difficult,a lightweight deep learning-based driver fatigue detection model,YOLO11-SFL,is proposed.The model introduces StarNet as the backbone network to simplify the overall network structure,thereby reducing the total number of parameters and computational costs while improving operational efficiency.A FasterNet Block is adopted to enhance the C3k2,optimizing its convolutional structure to eliminate redundant com-putations.Furthermore,a Lightweight Asymmetric Detection Head(LADH)with multi-level compression techniques is employed to improve the model's adaptability to detect objects of various sizes.This ap-proach reduces the number of model parameters while compensating for some of the accuracy loss associ-ated with lightweight models.Experimental results indicate that the proposed fatigue detection model,YOLO11-SFL,maintains high accuracy while significantly reducing its parameter count,computational load,and model size to 57.7%,57.1%,and 60.2%of those of the original YOLO11 model,respec-tively.Comparisons with other lightweight object detection models validate the effectiveness and superiority of this method in balancing lightweight design and performance.关键词
疲劳检测/深度学习/轻量化/非对称多级压缩技术Key words
fatigue detection/deep learning/lightweight/asymmetric multi-level compression technology分类
交通工程引用本文复制引用
张洋,朱泽德,李宸翔,王道斌..改进YOLO11的轻量化驾驶员疲劳检测算法[J].山东理工大学学报（自然科学版）,2026,40(5):15-21,7.
