报告时间:2019年8月28日(星期三)晚上20:00(北京时间)
主题:焦点方谈:物体和关键点检测
报告主持人:欧阳万里 (悉尼大学)
报告嘉宾:王井东 (Microsoft Research Asia )
报告题目:Deep High-Resolution Representation Learning for Visual Recognition
报告嘉宾:刘子纬 (The Chinese University of Hong Kong)
报告题目:The Glimpse of Detectrons: Dynamic Forwarding and Routing in Modern Detectors
Panel议题:
1. 如何看待近几年检测方面的发展?
2. 物体和关键点检测有哪些尚未解决的问题?
3. 物体和关键点检测有哪些值得关注的新问题?
4. 经典分割任务背景下,逐渐涌现出一些新鲜的任务,如模态实例分割。这些新任务会带来怎样的挑战与机遇?
5. 目前最先进的目标检测是否已经能用于自动驾驶?
6. 各比赛榜单(如COCO) SOTA 的刷新,主要驱动是什么?是更大算力的投入,更复杂的模型,更先进的技术,或是其他因素?
7. 相对于之前的feature engineering+DPM,基于深度学习的目标检测框架在性能方面带来了质的飞跃,这一领域的未来发展前景如何?
8. 有哪些值得关注的问题迫切需要解决?
9. One stage detector在未来是否会取代two stage detector?
Panel嘉宾:
王井东(Microsoft Research Asia )、刘子纬(The Chinese University of Hong Kong)、代季峰(Microsoft Research Asia)、卢策吾(上海交通大学)
*欢迎大家在下方留言提出主题相关问题,主持人和panel嘉宾会从中选择若干热度高的问题加入panel议题!
报告嘉宾:王井东(Microsoft Research Asia )
报告时间:2019年8月28日(星期三)晚上20:00(北京时间)
报告题目:Deep High-Resolution Representation Learning for Visual Recognition
报告人简介:
Jingdong Wang is a Senior Principal Research Manager with the Visual Computing Group, Microsoft Research, Beijing, China. His areas of current interest include neural architecture design, human pose estimation, semantic segmentation, large-scale indexing, and salient object detection. He has authored one book and 100+ papers in top conferences and prestigious international journals in computer vision, multimedia, and machine learning. He authored a comprehensive survey on learning to hash in TPAMI. His paper was selected into the Best Paper Finalist at ACM MM 2015. Dr. Wang has been an Associate Editor of IEEE TPAMI, IEEE TCSVT and IEEE TMM. He was an Area Chair or a Senior Program Committee Member of top conferences, such as CVPR, ICCV, ECCV, AAAI, IJCAI, and ACM Multimedia. He is an ACM Distinguished Member and a Fellow of the IAPR.
个人主页:
https://jingdongwang2017.github.io
报告摘要:
High-resolution representations are essential for position-sensitive vision problems, such as human pose estimation, semantic segmentation, and object detection. Existing state-of-the-art frameworks first encode the input image as a low-resolution representation through a subnetwork that is formed by connecting high-to-low resolution convolutions in series (e.g., ResNet, VGGNet), and then recover the high-resolution representation from the encoded low-resolution representation. Instead, our proposed network, named as High-Resolution Network (HRNet), maintains high-resolution representations through the whole process. There are two key characteristics: (i) Connect the high-to-low resolution convolution streams in parallel; (ii) Repeatedly exchange the information across resolutions. The benefit is that the resulting representation is semantically richer and spatially more precise. We show the superiority of the proposed HRNet in a wide range of applications, including human pose estimation, semantic segmentation, and object detection, suggesting that the HRNet is a stronger backbone for computer vision problems. All the codes are available at https://github.com/HRNet.
报告嘉宾:刘子纬(The Chinese University of Hong Kong)
报告时间:2019年8月28日(星期三)晚上20:30(北京时间)
报告题目:The Glimpse of Detectrons: Dynamic Forwarding and Routing in Modern Detectors
报告人简介:
Dr. Ziwei Liu is a research fellow at the Chinese University of Hong Kong. Prior to that, he served as a post-doctoral researcher at University of California, Berkeley, from 2017 to 2018. He received the Ph.D. degree from the Chinese University of Hong Kong (CUHK) in 2017, advised by Prof. Xiaoou Tang and Prof. Xiaogang Wang.
His research interest covers computer vision, machine learning and computer graphics. In recent years, he primarily focused on deep learning and its applications on open-world visual understanding and human sensing. He has published over 30 papers on top conferences and journals (e.g. CVPR, ICCV, ECCV, SIGGRAPH, T-PAMI and TOG) with more than 3,000 citations. He is the receipt of Microsoft Young Fellowship, Hong Kong PhD Fellowship and ICCV Young Researcher Award. He has contributed several vision benchmarks such as CelebA and DeepFashion. He has also won the championship of multiple international competitions including DAVIS 2017 and MSCOCO 2018.
个人主页:
https://liuziwei7.github.io/
报告摘要:
With substantial progress in the last five years, the focus of object detection research has shifted to tackle the complex scenarios such as massive categories and cluttered scenes. In this talk, I will summarize the recent advances in multimedia lab, from the perspective of dynamic forwarding and routing. Specifically, the talk will cover FishNet, guided anchoring and CARAFE, which enable dynamic spatial feature propagation during the feed-forward pass. The talk will also introduce a family of cascading models (chained cascade and hybrid task cascade) that incorporates the dynamic routing mechanism into modern detectors. I will conclude this talk by discussing some open questions and potential directions ahead.
Panel嘉宾:代季峰(Microsoft Research Asia)
嘉宾简介:
代季峰,于2009年和2014年分别获得清华大学自动化系本科和博士学位,2012年至2013年在加州大学洛杉矶分校访学,现任微软亚洲研究院视觉计算组资深研究员。代季峰博士的主要研究领域为计算机视觉中的物体检测、分割问题,和深度学习算法。他在顶级国际会议和期刊上共发表20余篇论文,据Google Scholar统计共计被引用4600余次。他曾经连续两年在本领域内权威的COCO物体识别竞赛中获得第一名。他是IJCV的编委,AAAI 2018的Senior Program Committee member.
个人主页:
https://jifengdai.org/
Panel嘉宾:卢策吾(上海交通大学)
嘉宾简介:
卢策吾,上海交通大学研究员,博士生导师,2016年获国家青年千人计划。2018年被《麻省理工科技评论》评选为中国“35位35以下科技创新青年”(MIT TR35)。2019年获求是杰出青年学者(全国所有学科10人)。加入交大前为斯坦福人工智能实验室博士后。目前担任CVM 2018大会主席,近几年发表(含接收)50多篇CCF A类文章,担任CVPR 2020 领域主席(Area Chair),CVM 2018 程序主席。研究方向,人体姿态与行为理解(代表作 alphapose,hake),机器人视觉。
个人主页:
http://mvig.sjtu.edu.cn/
主持人:欧阳万里(悉尼大学)
主持人简介:
欧阳万里于香港中文大学电子工程系获得博士学位。现任悉尼大学高级讲师(相当于美国体制副教授)。ICCV最佳审稿人,IJCV客座编辑(Guest Editor),IEEE高级会员。担任TPAMI, IJCV, TOG, TIP, CVPR, ICCV, SIGGRAPH等期刊/会议的审稿人。研究方向包括计算机视觉,模式识别,深度学习,图像处理等。主要从事基于深度学习的物体检测与跟踪,以及与人相关的问题的课题研究。作为一作在TPAMI和IJCV发表7篇文章。
个人主页:
https://wlouyang.github.io
19-21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“21期”,获取直播地址。
特别鸣谢本次Webinar主要组织者:
主办AC:欧阳万里(悉尼大学)
协办AC:徐畅(悉尼大学),樊彬 (中科院自动化所)
责任AC:王乃岩(图森未来)
VALSE Webinar改版说明:
自2019年1月起,VALSE Webinar改革活动形式,由过去每次一个讲者的方式改为两种可能的形式:
1)Webinar专题研讨:每次活动有一个研讨主题,先邀请两位主题相关的优秀讲者做专题报告(每人30分钟),随后邀请额外的2~3位嘉宾共同就研讨主题进行讨论(30分钟)。
2)Webinar特邀报告:每次活动邀请一位资深专家主讲,就其在自己熟悉领域的科研工作进行系统深入的介绍,报告时间50分钟,主持人与主讲人互动10分钟,自由问答10分钟。
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、H、I群已满,除讲者等嘉宾外,只能申请加入VALSE J群,群号:734872379);
*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四发布下一周Webinar报告的通知及直播链接。
8、Webinar报告的PPT(经讲者允许后),会在VALSE官网每期报告通知的最下方更新[slides]。
9、Webinar报告的视频(经讲者允许后),会更新在VALSE爱奇艺空间,请在爱奇艺关注Valse Webinar进行观看。