本文提出基于MobileNetV3-CBAM+SSD的人脸检测模型,首先将MobileNetV3轻量化网络代替SSD 模型的主干特征提取网络VGG-16,提高了模型的检测速度,然后在SSD模型中引入CBAM轻量级注意力机制,提高了模型的检测精度,最后将本文所提算法与SSD和MobileNetV3-SSD算法进行实验性能对比。实验结果表明,本文提出的人脸检测模型在DataSet数据集下平均精度均值达到94.58%,提高了9.91%,检测速度提高了42帧/s,计算参数和模型大小减少,基本满足应用要求。
Abstract
In the paper,a face detection model based on MobileNetV3-CBAM+SSD is proposed.Firstly,the MobileNetV3 lightweight network is used to replace the backbone feature extraction network VGG-16 of SSD model,which improves the detection speed of the model.Then,the lightweight attention mechanism of CBAM is introduced into the SSD model,which improves the detection accuracy of the model.Finally,the experimental performance of the proposed algorithm is compared with SSD and MobileNetV3-SSD algorithm.The results show that the average accuracy of the face detection model proposed in this paper under the DataSet data set is 94.58%,which is increased by 9.91%,the detection speed is increased by 42 frames/s,and the calculation parameters and model size are reduced,which basically meets the application requirements.
关键词
SSD模型 /
MobileNetV3网络 /
CBAM注意力机制 /
人脸检测
Key words
SSD model /
MobileNetV3 network /
CBAM attention mechanism /
face detection
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 徐冬阳.人脸检测与识别技术的发展及应用[J].电子世界,2018(1):193-194.
[2] Hasan Md Khaled,Ahsan MdShamim,AbdullahAlMamun,et al.Human Face Detection Techniques:A Comprehensive Review and Future Research Directions[J].Electronics,2021,10(19).
[3] Zheng Guangyong,Xu Yuming.Efficient face detection and tracking in video sequences based on deep learning[J].Information Sciences,2021(568).
[4] Tao Shifan,Li Yufeng,Huang Yufeng,et al.Face Detection Algorithm Based on Deep Residual Network[J].Journal of Physics:Conference Series,2021,1802(3).
[5] 黄绍欣,田秀云.基于深度学习的人脸检测算法研究[J].机电工程技术,2022,51(1):126-129.
[6] 杨璐,吴陈.基于SSD算法的人脸目标检测的研究[J].计算机技术与发展,2019,29(10):181-185.
[7] 王嘉珺.基于SSD卷积网络的人脸检测算法研究[D].厦门:厦门大学,2018.
[8] 陈力.基于改进型SSD网络的人脸检测[D].南昌:南昌大学,2020.
[9] 张骏,朱标,吉涛.基于MobileNet-SSD的红外人脸检测算法[J].激光与红外,2021,51(1):107-113.
[10] 王麟阁,蒋宝军,潘铁军.基于层级注意力增进网络的多尺寸遮挡人脸检测[J].数据采集与处理,2022,37(1):73-81.
[11] 李保华,王海星.基于增强卷积神经网络的尺度不变人脸检测方法[J].红外与激光工程,2022,11(4):1-8.
[12] 钟志峰,夏一帆,周冬平,等.基于改进YOLOv4的轻量化目标检测算法[J].计算机应用,2022,11(4):1-8.
[13] 郝帅,张旭,马旭,等.基于CBAM-YOLOv5的煤矿输送带异物检测[J].煤炭学报,2022(2):1-11.
[14] 赵会盼,乔任重,张学军.一种基于嵌入式平台的轻量化目标检测识别方法[J].计算机与网络,2021,47(15):55-60.
[15] 张运波,易鹏飞,周东生,等.深度可分离卷积和标准卷积相结合的高效行人检测器[J].图学学报,2022,11(4):1-10.
[16] Fu Huixuan,Song Guoqing,Wang Yuchao.Improved YOLOv4 Marine Target Detection Combined with CBAM[J]. Symmetry,2021,13(4).
[17] Fu Huixuan,Song Guoqing,Wang Yuchao.Improved YOLOv4 Marine Target Detection Combined with CBAM[J]. Symmetry,2021,13(4).
[18] Wenhao Cao,Zhuoyu Feng,Dongyao Zhang,et al.Facial Expression Recognition via a CBAM Embedded Network[J].Procedia Computer Science,2020(174):463-477.
基金
*陕西省技术创新引导专项(基金)计划(2019CGXNG-015)。
PDF(3401 KB)
