20180425-10 谢晋：Deep learning based 3D shape representation

报告嘉宾：谢晋（南京理工大学计算机学院）

报告时间：2018年04月25日（星期三）晚上20:00（北京时间）

报告题目：Deep learning based 3D shape representation

主持人：张林（同济大学）

报告人简介：

Jin Xie received his Ph.D. degree from the Department of Computing, The Hong Kong Polytechnic University. He is a professor at Nanjing University of Science and Technology (NJUST), China. Prior to joining NJUST, he was a research scientist at New York University Abu Dhabi and New York University Tandon School of Engineering. His research interests include image forensics, computer vision and machine learning. Currently he is focusing on 3D computer vision with convex optimization and deep learning methods, including 3D shape analysis, 3D object detection and 3D scene understanding. He has published papers in top conferences and journals, including CVPR, ECCV, AAAI, ACM MM, IEEE TPAMI and TIP. He has served as a PC member for CVPR, ICCV, ECCV, ACM MM, ICPR and ACPR, a journal reviewer for IEEE TPAMI, TIP, TNNLS, TMM, TCYB, TCSVT, PR and PRL. He was a special issue chair for ACPR 2017 and a guest editor for Pattern Recognition.

讲者个人主页：

https://csjin.github.io/index.html

相关文献： 

1.Learning Barycentric representations of 3D shapes for sketch-based 3D shape retrieval， Jin Xie, Guoxian Dai, Fan Zhu and Yi Fang，CVPR, 2017.

2.Learned binary spectral shape descriptor for 3D shape correspondence, Jin Xie, Meng Wang and Yi Fang, CVPR, 2016.

3.Heat diffusion long-short term memory learning for 3D shape analysis, Fan Zhu, Jin Xie and Yi Fang, ECCV, 2016.

4.Deepshape:deep learned shape descriptor for 3D shape matching and retrieval, Jin Xie, Yi Fang, Fan Zhu and Edward K.Wong, CVPR, 2015.

报告摘要：

Advances in deep learning via deep neural networks have resulted in great gains in the computer vision community. Different from 2D images, 3D shapes do not contain rich textures and colors, but contain geometric structures. How to represent 3D shapes with deep neural networks is a challenging problem. In this talk, I will present deep learning based 3D shape representations and their applications in 3D shape analysis. First, I will introduce the heat diffusion theory in 3D shape analysis. Based on the heat diffusion theory, I will then present 3D shape feature extraction with deep neural networks for 3D shape retrieval and correspondence. In addition, the potential applications of 3D shape representations will also be discussed.

特别鸣谢本次Webinar主要组织者：

VOOC责任委员：张林（同济大学）

VODB协调理事：张兆翔（中科院自动化所）

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），直播链接会在报告当天（每周三）在VALSE微信公众号和VALSE QQ群发布；

*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10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，请勿送花、打赏等，也不要说无关话语，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。

[slides]