报告嘉宾:陶文兵(华中科技大学) 报告时间:2018年05月30日(星期三)晚上20:00(北京时间) 报告题目:Efficient Large-Scale 3D Reconstruction 主持人:杜博(武汉大学) 报告人简介: 陶文兵,教授,博士生导师,2004年12月在华中科技大学图像所获模式识别与智能系统专业工学博士学位,2013年11月晋升为教授,现为华中科技大学自动化学院图像识别与人工智能研究所/多谱信息处理技术国家级重点实验室/图像信息与智能控制教育部重点实验室教授/博士生导师。主要研究方向为图像分割,目标检测跟踪,大尺度三维重建,机器人目标识别与获取等领域。在国内外学术期刊和国际会议上共发表论文80多篇,SCI收录60余篇。2014年——2017年连续四年入选中国高被引学者(Most Cited Chinese Researchers)计算机领域。授权及申请发明专利30余项。 个人主页: http://grid.hust.edu.cn/wenbingtao/ 报告摘要: 3D reconstruction plays an important role in most computer vision. In this talk, we will introduce three core steps of the whole 3D reconstruction pipeline: GPU accelerated large scale image matching, large scale structure-from-motion, and multi-view stereo for 3D dense reconstruction. At first, we present a GPU-based cascade hashing strategy to greatly accelerate image matching. Secondly, we will introduce a trilaminar multiway reconstruction tree algorithm for large-scale structure-from-motion. The trilaminar multiway reconstruction tree divides the image set into clusters suitable for reconstruction as well as finds multiple reliable and stable starting points. This promises that the scene is always reconstructed from dense places to sparse areas, which can reduce error accumulation when images have weak overlap. Last, we present a multi-view stereo method with asymmetric checkerboard propagation and multi-hypothesis joint view selection. The method exploits the asymmetric checkerboard propagation to make full use of massive parallel computing power of GPU and spread reasonable hypotheses further. To robustly aggregate matching costs, multi-hypothesis joint view selection is used to constructs a cost matrix and heuristically infer an aggregation view subset for each pixel. As a result, our method can run much faster than competing methods and achieve the sate-of-the-art. 参考文献: [1] Tao Xu, Kun Sun and Wenbing Tao*, GPU Accelerated Cascade Hashing Image Matching for Large Scale 3D reconstruction, arXiv. [2] Kun Sun, Wenbing Tao*, Multiple Starting Points Selection and Data Partitioning for Accurate, Efficient Structure from Motion, arXiv. [3] Qingshan Xu, Wenbing Tao*, Multi-View Stereo with Asymmetric Checkerboard Propagation and Multi-Hypothesis Joint View Selection, arXiv. 18-15期VALSE在线学术报告参与方式: 长按或扫描下方二维码,关注”VALSE“微信公众号(valse_wechat),后台回复”15期“,获取直播地址。 特别鸣谢本次Webinar主要组织者: VOOC责任委员:杜博(武汉大学) VODB协调理事:林倞(中山大学) 活动参与方式: 1、VALSE Webinar活动依托在线直播平台进行,活动时讲者会上传PPT或共享屏幕,听众可以看到Slides,听到讲者的语音,并通过聊天功能与讲者交互; 2、为参加活动,请关注VALSE微信公众号:valse_wechat 或加入VALSE QQ群(目前A、B、C、D、E、F、G群已满,除讲者等嘉宾外,只能申请加入VALSE H群,群号:701662399); *注:申请加入VALSE QQ群时需验证姓名、单位和身份,缺一不可。入群后,请实名,姓名身份单位。身份:学校及科研单位人员T;企业研发I;博士D;硕士M。 3、在活动开始前5分钟左右,讲者会开启直播,听众点击直播链接即可参加活动,支持安装Windows系统的电脑、MAC电脑、手机等设备; 4、活动过程中,请不要说无关话语,以免影响活动正常进行; 5、活动过程中,如出现听不到或看不到视频等问题,建议退出再重新进入,一般都能解决问题; 6、建议务必在速度较快的网络上参加活动,优先采用有线网络连接; 7、VALSE微信公众号会在每周一推送上一周Webinar报告的总结及视频(经讲者允许后),每周四发布下一周Webinar报告的通知及直播链接。 |
小黑屋|手机版|Archiver|Vision And Learning SEminar
GMT+8, 2024-11-24 08:24 , Processed in 0.013033 second(s), 15 queries .
Powered by Discuz! X3.4
Copyright © 2001-2020, Tencent Cloud.