GAMES Webinar 2019 – 82期(CVPR 2018三维视觉论文报告)| 祁芮中台(Facebook AI Research),苏航(马萨诸塞大学安姆斯特分校)

【GAMES Webinar 2019-82期(CVPR 2018三维视觉论文报告)】
报告嘉宾1:祁芮中台,Facebook AI Research
报告时间:2019年1月17日 晚8:00-8:45(北京时间)
主持人:徐凯,国防科技大学(个人主页:http://www.kevinkaixu.net
报告题目:Frustum PointNets for 3D Object Detection from RGB-D Data
报告摘要:
In this talk, we introduce a new method for 3D object detection from RGB-D data in both indoor and outdoor scenes. While previous methods focus on images or 3D voxels, often obscuring natural 3D patterns and invariances of 3D data, we directly operate on raw point clouds by popping up RGB-D scans. However, a key challenge of this approach is how to efficiently localize objects in point clouds of large-scale scenes (region proposal). Instead of solely relying on 3D proposals, our method leverages both mature 2D object detectors and advanced 3D deep learning for object localization, achieving efficiency as well as high recall for even small objects. Benefited from learning directly in raw point clouds, our method is also able to precisely estimate 3D bounding boxes even under strong occlusion or with very sparse points. Evaluated on KITTI and SUN RGB-D 3D detection benchmarks, our method outperforms the state of the art by remarkable margins while having real-time capability.
Project website: http://stanford.edu/~rqi/frustum-pointnets/
讲者简介:
Charles Ruizhongtai Qi is currently a postdoctoral researcher at Facebook AI Research, working with Leonidas Guibas and Jitendra Malik. He received his Ph.D. from Stanford University (AI Lab and Geometric Computation Group) in 2018. Prior to joining Stanford in 2013, he got his B.E. degree from Tsinghua University in Electronic Engineering, with an outstanding graduate award. His research focuses on computer vision and deep learning, specifically on 3D deep learning, 3D object recognition and using 3D shape priors for image understanding.
讲者个人主页:http://charlesrqi.com

 

报告嘉宾2:苏航,马萨诸塞大学安姆斯特分校
报告时间:2019年1月17日 晚8:45-9:30(北京时间)
主持人:徐凯,国防科技大学(个人主页:http://www.kevinkaixu.net
报告题目:基于多视角图像与点云的三维特征学习
报告摘要:
在这次报告中,我将主要介绍在三维特征学习方面我的两项工作以及正在进行的一些相关研究。计算机视觉上长期存在关于三维描述与二维多视角描述之间的争论。多视角卷积神经网络(MVCNN)试图从三维物体识别角度对此进行探讨。 尽管最近出现很多基于三维描述的特征学习工作,我们发现MVCNN,以及它近期的一些延伸工作,仍在大多数评测数据集上表现出很强的竞争力。稀疏网格网络(SPLATNet)是一个可以不需要转换,直接对三维点云进行操作的神经网络。除了其在点云处理上的较好性能,SPLATNet的另一优势是它能高效地整合点云和图像数据,结合三维描述与二维多视角描述各自的优点。最后,针对MVCNN和SPLATNet各自的优势及局限,我将对近期我们实验室以及其他学者的相关工作进行简单的探讨。
讲者简介:
苏航目前就读于UMass-Amherst计算机视觉实验室,在导师Prof. Erik Learned-Miller与Prof. Subhransu Maji指导下攻读博士学位。在此之前他分别于北京大学与布朗大学取得本科及硕士学位,并曾在微软研究院及英伟达研究院实习。苏航的主要研究方向包括计算机视觉与图形学,并侧重于结合三维与二维信息的视觉特征学习方法。他关于SPLATNet的工作获得了CVPR’18 Best Paper Honorable Mention以及NVAIL Pioneering Research Award奖项。
讲者个人主页:http://cs.umass.edu/~hsu

 

GAMES主页的“使用教程”中有 “如何观看GAMES Webinar直播?”及“如何加入GAMES微信群?”的信息;
GAMES主页的“资源分享”有往届的直播讲座的视频及PPT等。

 

观看直播的链接:http://webinar.games-cn.org

You may also like...