Title:
Two Clouds of Points: The Problem at the Heart of Spatial Intelligence; Why the obvious distance is wrong — and how to fix it
Abstract:
Spatial intelligence — knowing what is where and recognizing the same thing seen twice — underpins robotics, navigation, and 3D scene understanding. Almost all of it reduces to one primitive: matching one unordered set of points to another. This talk explains why the standard way of measuring that match quietly biases everything built on top of it, and how ideas from probability, curved geometry, and information theory fix it without sacrificing efficiency. No prior background in 3D vision is assumed.
Speaker:
Ziming Zhang
Associate Professor, ECE Department, WPI
Bio:
Dr. Ziming Zhang is an Associate Professor at Worcester Polytechnic Institute. Before joining WPI he was a research scientist at Mitsubishi Electric Research Laboratories (MERL) in 2016-2019. Prior to that, he was a research assistant professor at Boston University. Dr. Zhang received his PhD in 2013 from Oxford Brookes University, UK, under the supervision of Prof. Philip H. S. Torr (now in the University of Oxford). His research areas lie in computer vision and machine learning, especially in object recognition/detection, data-efficient learning (e.g. zero-shot learning) and applications (e.g. person re-identification), deep learning, optimization. His works have appeared in PAMI, CVPR, ICCV, ECCV, NIPS. He serves as a review/PC member for top conferences (e.g. CVPR, ICCV, NIPS, ICML, ICLR, AAAI, AISTATS, IJCAI) and journals (e.g. PAMI, IJCV, JMLR). He won the R&D100 Award 2018.
Host: Professor John McNeill