首页 | 本学科首页   官方微博 | 高级检索  
     

基于轻量级卷积神经网络的人脸识别方法
引用本文:黄良辉,康祖超,张昌凡,程 涛.基于轻量级卷积神经网络的人脸识别方法[J].湖南工业大学学报,2019,33(2):43-47.
作者姓名:黄良辉  康祖超  张昌凡  程 涛
作者单位:广东南海鹰视通达科技有限公司,湖南工业大学 电气与信息工程学院,湖南工业大学 电气与信息工程学院,湖南工业大学 电气与信息工程学院
基金项目:国家自然科学基金资助项目(61773159)
摘    要:针对传统卷积神经网络在人脸识别中模型复杂程度高、处理数据较慢的问题,提出一种轻量级卷积神经网络算法。首先,通过对数据集采用剪裁、旋转等方式增强样本数据;然后,采用基于MobileNet的轻量级卷积神经网络对样本数据进行特征提取,并采用SSD目标检测器对样本数据中的人脸进行识别;最后,利用Python编程实现上述算法,并与传统的人脸识别算法进行比较。实验结果表明,采用的轻量级卷积神经网络算法在不失精度的前提下,处理速度更快,模型复杂程度更低。

关 键 词:轻量级卷积神经网络  MobileNet  目标检测  人脸识别
收稿时间:2018/11/11 0:00:00

Research on Face Recognition Technology Based on Lightweight Convolutional Neural Networks
HUANG Lianghui,KANG Zuchao,ZHANG Changfan and CHENG Tao.Research on Face Recognition Technology Based on Lightweight Convolutional Neural Networks[J].Journal of Hnnnan University of Technology,2019,33(2):43-47.
Authors:HUANG Lianghui  KANG Zuchao  ZHANG Changfan and CHENG Tao
Abstract:n view of the flaws of high complexity and slow data processing of face recognition found in traditional convolution neural networks, a lightweight convolutional neural network algorithm has thus been proposed. Firstly, the sample data is to be enhanced by clipping and rotating the data set. Then, a lightweight convolutional neural network based on MobileNet is used to extract the features from sample data, and an SSD target detector is used for face recognition in sample data. Finally, the above algorithm is implemented by Python programming, followed by a comparison with the outcome of the traditional face recognition algorithm. The experimental results show that the proposed lightweight convolution neural network algorithm is faster in the processing speed with a lower model complexity and retained accuracy.
Keywords:
本文献已被 CNKI 等数据库收录!
点击此处可从《湖南工业大学学报》浏览原始摘要信息
点击此处可从《湖南工业大学学报》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号