首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于深度卷积网络与空洞卷积融合的人群计数
引用本文:盛馨心,苏颖,汪洋.基于深度卷积网络与空洞卷积融合的人群计数[J].上海师范大学学报(自然科学版),2019,48(5):479-484.
作者姓名:盛馨心  苏颖  汪洋
作者单位:上海师范大学 信息与机电工程学院, 上海 200234,上海师范大学 信息与机电工程学院, 上海 200234,上海师范大学 信息与机电工程学院, 上海 200234
摘    要:利用空洞卷积设置不同空洞率,得到不同感受野的特点,提出一种基于深度卷积Visual Geometry Group19(VGG19)和空洞卷积相融合的结构.所采用的结构不受输入图像尺寸以及分辨率影响,通过设置锯齿状空洞率,扩大网络的感受野,在保持分辨率良好的情况下,可以较为精确地定位目标,提高检测准确性.经验证,该算法在Shanghai-tech标准数据集上具有较高的实验准确率.

关 键 词:人群计数  Visual  Geometry  Group19(VGG19)  空洞卷积  Shanghai-tech数据集
收稿时间:2019/7/27 0:00:00

Crowd counting based on fusion of deep convolutional network and dilated convolution
SHENG Xinxin,SU Ying and WANG Yang.Crowd counting based on fusion of deep convolutional network and dilated convolution[J].Journal of Shanghai Normal University(Natural Sciences),2019,48(5):479-484.
Authors:SHENG Xinxin  SU Ying and WANG Yang
Institution:College of Information, Mechanical and Electrical Engineering, Shanghai Normal University, Shanghai 200234, China,College of Information, Mechanical and Electrical Engineering, Shanghai Normal University, Shanghai 200234, China and College of Information, Mechanical and Electrical Engineering, Shanghai Normal University, Shanghai 200234, China
Abstract:A combined structure based on Visual Geometry Group19(VGG19) and dilated convolution with different receptive field was proposed for high density crowd counting in the paper.The structure adopted would not be affected by the size and resolution of the input image.By setting the serration dilation rate,the network receptive field was expanded,and the target could be accurately localized without any loss of resolution,which improved the accuracy of detection.Finally,the experimental results showed that the algorithm had higher accuracy on the standard data set of Shanghai-tech.
Keywords:crowd counting  Visual Geometry Group19(VGG19)  dilated convolution  Shanghai-tech data set
本文献已被 CNKI 等数据库收录!
点击此处可从《上海师范大学学报(自然科学版)》浏览原始摘要信息
点击此处可从《上海师范大学学报(自然科学版)》下载免费的PDF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