软件导刊2025，Vol.24Issue(8)：59-64,6.DOI:10.11907/rjdk.241602
集成人脸关键点的轻量化YOLOv7人脸检测算法
Lightweight YOLOv7 Face Detection Algorithm Integrated with Facial Keypoints
摘要
Abstract
Facial detection,as an important component of computer vision technology,has been widely applied in fields such as fraud preven-tion,sentiment analysis,and age prediction.However,with the popularity of mobile devices and smart cameras,the acquisition of image and video data has become more convenient,but the computing power of devices is limited,and there is an urgent need for faster and more effi-cient face detection algorithms.To this end,a lightweight YOLOv7 face detection algorithm integrating facial keypoints is proposed,aiming to achieve efficient face detection and adapt to mobile and embedded devices with limited computing power.The algorithm first uses the GhostNet V2 lightweight module with higher computational efficiency to replace the ELAN module of YOLOv7,reducing the model parameter count by 63.9%;Secondly,the improved PFLD facial keypoint detection algorithm will be integrated into the Head section of YOLOv7 to address is-sues such as diverse facial sample poses,inconsistent image quality,and uneven sample distribution in the dataset,forming a multi task fa-cial detection network with facial detection as the main focus and facial keypoint detection as a secondary approach.The experimental results on the WIDER FACE dataset show that the proposed algorithm achieves average accuracy of 95.9%,93.5%,and 83.5%on the Easy,Medi-um,and Hard subsets,respectively.Compared with the original YOLOv7 model,it improves by 3.8%,2.9%,and 4.1%,respectively,while reducing the model parameter count by 41%.It is a fast and efficient face detection algorithm.关键词
人脸检测/轻量化算法/YOLOv7/GhostNet V2/人脸关键点Key words
face detection/lightweight algorithm/YOLOv7/GhostNet V2/facial keypoints分类
信息技术与安全科学引用本文复制引用
曹一波,胡伊斐,刘顺..集成人脸关键点的轻量化YOLOv7人脸检测算法[J].软件导刊,2025,24(8):59-64,6.基金项目
宁波市重点技术研发项目(2021E007) （2021E007）
