PHD: A Deep Learning Based Human Detection Framework for Panoramic Videos

Jinting Tang, Zhenhui Chen, Yongkai Huo, Peichang Zhang
{"title":"PHD: A Deep Learning Based Human Detection Framework for Panoramic Videos","authors":"Jinting Tang, Zhenhui Chen, Yongkai Huo, Peichang Zhang","doi":"10.1109/WCSP.2019.8928143","DOIUrl":null,"url":null,"abstract":"Panoramic video has attracted substantial research attention as the coming video format. It is capable of providing 360 degree immersive experience of omnidirectional visual information. State-of-the-art detection networks may fail to detect humans on spherical images, which are normally represented in deformed rectangular shapes. In this paper, we propose a socalled Panoramic Human Detection (PHD) scheme to address the task of human detection in panoramic videos. Moreover, the PHD method is designed to detect humans by extracting multiple overlapping sub-images from each integral spherical image, where three-dimensional rotation of spherical images is employed to ensure consistency of sub-images. Two detection box filters are designed for removing redundant boxes. Our PHD method is capable of accomplishing the task of human detection in various panoramic video types. Experiments prove that our PHD method outperforms the baseline by 35% and 48.6% in terms of precision and recall, respectively.","PeriodicalId":108635,"journal":{"name":"2019 11th International Conference on Wireless Communications and Signal Processing (WCSP)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2019 11th International Conference on Wireless Communications and Signal Processing (WCSP)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/WCSP.2019.8928143","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 2

Abstract

Panoramic video has attracted substantial research attention as the coming video format. It is capable of providing 360 degree immersive experience of omnidirectional visual information. State-of-the-art detection networks may fail to detect humans on spherical images, which are normally represented in deformed rectangular shapes. In this paper, we propose a socalled Panoramic Human Detection (PHD) scheme to address the task of human detection in panoramic videos. Moreover, the PHD method is designed to detect humans by extracting multiple overlapping sub-images from each integral spherical image, where three-dimensional rotation of spherical images is employed to ensure consistency of sub-images. Two detection box filters are designed for removing redundant boxes. Our PHD method is capable of accomplishing the task of human detection in various panoramic video types. Experiments prove that our PHD method outperforms the baseline by 35% and 48.6% in terms of precision and recall, respectively.
博士:基于深度学习的全景视频人体检测框架
全景视频作为一种新兴的视频格式已经引起了广泛的研究。能够提供360度全方位的视觉信息沉浸式体验。最先进的检测网络可能无法在球形图像上检测到人类,这些图像通常以变形的矩形形状表示。在本文中,我们提出了一个所谓的全景人体检测(PHD)方案来解决全景视频中的人体检测任务。此外,PHD方法通过从每个积分球面图像中提取多个重叠的子图像来检测人,其中利用球面图像的三维旋转来保证子图像的一致性。两个检测盒过滤器设计用于去除多余的盒子。我们的PHD方法能够在各种全景视频类型中完成人体检测的任务。实验证明,我们的PHD方法在准确率和召回率方面分别优于基线35%和48.6%。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信