20200909-22 图像与视频分割：观像素之微，察品类之盛

报告时间:2020年09月09日(星期三)晚上20:00(北京时间)

主题:图像与视频分割：观像素之微，察品类之盛

主持人:李冠彬(中山大学)

报告嘉宾：魏云超(University of Technology Sydney)

报告题目：Recent Progresses in Visual Segmentation

报告嘉宾：王文冠(ETH Zurich)

报告题目：Pixel-Wise Pattern Understanding from Structural Visual Data

Panel嘉宾：

沈春华(University of Adelaide)、凌海滨(Stony Brook University)、沈建冰(Inception Institute of Artificial Intelligence)、胡瀚(MSRA)、魏云超(University of Technology Sydney)、王文冠(ETH Zurich)

Panel议题：

1. 图像分割的发展当前面临哪些问题？哪些方面有望在近几年取得突破？

2. 视频分割与图像分割有何异同？有哪些实际应用？未来的发展方向及面临的主要困难是？

3. 当前有人认为语义分割等方向性能提升困难、呈现内卷化和进入瓶颈期的趋势，如何看待这一观点？对新入门的同学有何建议？

4. 视频分割和目标跟踪的区别和联系是什么？这两个方向有没有可能统一？

5. 从样本数量、代表性和均衡性来看，当前的图像语义分割和视频分割数据集是否合理？后续数据集的构建和扩展需要考虑什么因素才能更好的推进领域的研究？

*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：魏云超(University of Technology Sydney)

报告时间：2020年09月09日(星期三)晚上20:00(北京时间)

报告题目：Recent Progresses in Visual Segmentation

报告人简介：

魏云超，悉尼科技大学助理教授。曾为伊利诺伊大学香槟分校、新加坡国立大学博后，2016年于北京交通大学获得博士学位。曾获2019年澳大利亚研究理事会青年研究奖、2019年中国中国图象图形学学会科学技术奖一等奖、2016年中国电子学会优秀博士论文奖，并多次获得ImageNet、人体精细化分割(LIP)、视频物体分割(DAVIS)等国际竞赛的冠亚军。主要从事计算机视觉领域的相关研究，包括图像分类、视频/图像的物体检测和分割、弱监督/半监督学习等。

个人主页：

https://weiyc.github.io/

报告摘要：

Image/video segmentation is one of fundamental research areas in the computer vision community and has many potential real-world applications such as autonomous vehicle, robotics, and video editing. In this talk, Dr. Wei will present some of his recent works in image semantic segmentation, interactive object segmentation and video object segmentation.

参考文献：

[1] CCNet: Criss-Cross Attention for Semantic Segmentation, Zilong Huang, Xinggang Wang, Yunchao Wei, Lichao Huang, Humphrey Shi, Wenyu Liu, Thomas S. Huang, IEEE TPAMI, 2020.

[2] Interactive Object Segmentation with Inside-Outside Guidance, Shiyin Zhang, Jun Hao Liew, Yunchao Wei, Shikui Wei, Yao Zhao, IEEE CVPR, 2020.(oral)

[3] Collaborative Video Object Segmentation by Foreground-Background Integration, Zongxin Yang, Yunchao Wei, Yi Yang, ECCV, 2020.(spotlight)

报告嘉宾：王文冠(ETH Zurich)

报告时间：2020年09月09日(星期三)晚上20:30(北京时间)

报告题目：Pixel-Wise Pattern Understanding from Structural Visual Data

报告人简介：

王文冠，苏黎世联邦理工学院博后研究员。2018年博士毕业于北京理工大学，2016至2018年在加州大学洛杉矶分校(UCLA)访学。2018至2019年，在起源人工智能研究院(IIAI)任Senior Scientist。曾获百度奖学金、ACM 中国优博奖、中国人工智能学会优博奖、世界人工智能大会青年优秀论文奖等。带队在以下国际比赛的若干赛道获得先进名次：CVPR 2020 LID 和ICCV 2019 PIC 冠军，CVPR 2020 DAVIS 和CVPR 2020 LIP亚军，及CVPR 2020 Embodied-AI Habitat、CVPR 2019 WAD 和CVPR 2019 LIP 季军。主要研究方向为图像/视频分析、以人为中心的视觉理解、点云分割/检测、实体化AI等。

个人主页：

https://sites.google.com/view/wenguanwang

报告摘要：

As we live in a highly structured world, it is essential to explore and understand visual structured information in computer vision tasks, including image/video segmentation. For example, things that are semantically related are typically presented in a similar way, evidenced as the inter- and intra-image context, which plays a crucial role in semantic segmentation. Things also undergo continuous variations over time; in a video clip, the contents of different frames are often correlated, providing extra cues for video segmentation. Hence, things are composed of elements, which are interacted and organized in certain ways. Thus it is desired to understand semantics from a structured view, such as modeling human beings over multiple granularities--face parsing, pose estimation as well as fine-grained part segmentation. In this talk, I will advocate the value of structured information in image/video segmentation. As examples, I will present a line of my recent works on video object segmentation, human semantic parsing, and weakly supervised semantic segmentation.

参考文献：

[1] Zero-shot video object segmentation via attentive graph neural networks, Wenguan Wang, Xiankai Lu, et al, ICCV, 2019.(Oral)

[2] Hierarchical human parsing with typed part-relation reasoning, Wenguan Wang, Hailong Zhu, et al, CVPR, 2020.

[3] Mining cross-image semantics for weakly supervised semantic segmentation, Guolei Sun, Wenguan Wang, et al, ECCV, 2020.(Oral)

Panel嘉宾：沈春华(University of Adelaide)

嘉宾简介：

He is a Professor of Computer Science at University of Adelaide. He was awarded an ARC Future Fellowship in 2012. He is an Adjunct Professor of Data Science and AI at Faculty of Information Technology, Monash University. His research interests include Statistical Machine Learning and Computer Vision.

个人主页：

https://cs.adelaide.edu.au/~chhshen/

Panel嘉宾：凌海滨(Stony Brook University)

嘉宾简介：

He is now a SUNY Empire Innovation Professor in the Department of Computer Science of Stony Brook University. His research interests include computer vision, augmented reality, medical image analysis, visual privacy protection, and human computer interaction. He received Best Student Paper Award of ACM UIST in 2003 and NSF CAREER Award in 2014.

个人主页：

https://www3.cs.stonybrook.edu/~hling/

Panel嘉宾：沈建冰(Inception Institute of Artificial Intelligence)

嘉宾简介：

He is currently acting as the Lead Scientist with the Inception Institute of Artificial Intelligence, UAE. He is also a Full Professor with the School of Computer Science, Beijing Institute of Technology. He has published about 100 journal and conference papers such as IEEE TPAMI, CVPR, and ICCV. He has obtained many honors including the Fok Ying Tung Education Foundation from Ministry of Education, the Program for Beijing Excellent Youth Talents from Beijing Municipal Education Commission, and the Program for New Century Excellent Talents from Ministry of Education. His research interests include computer vision and deep learning. He is an Associate Editor of IEEE TIP, IEEE TNNLS and other journals.

个人主页：

https://scholar.google.com/citations?user=Q3NTToAAAAJ

Panel嘉宾：胡瀚(MSRA)

嘉宾简介：

胡瀚现任微软亚洲研究院视觉计算组主管研究员， 2008年和2014年在清华大学自动化系分别获得本科和博士学位，2016年获中国人工智能学会优秀博士论文奖。于2012年在宾夕法尼亚大学GRASP实验室做访问研究，加入微软亚洲研究院前曾在百度深度学习实验室工作。目前主要研究兴趣是视觉表征学习，视觉语言联合表征学习，以及视觉物体识别等等。将担任CVPR 2021领域主席。

个人主页：

https://ancientmooner.github.io/

主持人：李冠彬(中山大学)

主持人简介：

李冠彬，中山大学数据科学与计算机学院副教授，2016年获得香港大学博士学位。主要研究领域包括计算机视觉与机器学习，迄今为止累计发表论文80余篇，其中包含CCF A类/中科院一区论文52篇，包括IEEE TPAMI，TIP，TNNLS，TCYB等顶级期刊和CVPR，ICCV，ECCV，ICML，AAAI，IJCAI等顶级学术会议，Google Scholar引用超过2500次。曾获得ICCV 2019最佳论文提名奖、中国图象图形学会科学技术一等奖（第三完成人）、ACM中国新星提名奖等荣誉。主持了包括广东省杰出青年基金、国家自然科学基金面上项目、国家自然科学基金青年项目、CCF-腾讯犀牛鸟科研基金等10多项科研项目。担任The Visual Computer编委，TPAMI、TIP、TNNLS、TMM、TCYB、TOG等权威期刊的审稿人，CVPR、ICCV、IJCAI、AAAI等国际会议程序委员会委员。

个人主页：

http://guanbinli.com/

20-22期VALSE在线学术报告参与方式：

扫描下方二维码，关注“VALSE”微信公众号 (valse_wechat)，后台回复“22期”，获取直播地址。

特别鸣谢本次Webinar主要组织者：

主办AC：王文冠(ETH Zurich)

协办AC：魏云超(University of Technology Sydney)、李冠彬(中山大学)

责任AC：王兴刚(华中科技大学)