Real-time People Tracking, Human Pose Estimation and Action Recognition Using 4D Computer Vision
Understanding people's movement and activity in crowds and clutter is a challenging problem. We proposal real-time 4D computer vision methods for robust people tracking, pose estimation and action recognition. With multiple depth sensors, we can reconstruct a dynamic point cloud of the scene. Our method works on such 4D data. Our method greatly outperforms traditional approaches and at the same time is much faster.
Research Scientist, Facebook
Dr. Hao Jiang is a Research Scientist at Facebook Reality Labs. He received PhD in computer science from Simon Fraser University in 2006 and spent a year from 2006 to 2007 as a Postdoctoral Research Fellow at the University of British Columbia. From 2007 to 2017. He was an Assistant Professor and then a Tenured Associate Professor in the Computer Science Department at Boston College. His research spans human pose, tracking, action understanding, 3D computer vision, egocentric vision and deep learning. He developed the people tracking and object association algorithm for Amazon Go, when he was a Senior Research Scientist at Amazon in 2013-2014. In 2016, he led the Computer Vision Team at AutoX.ai at Silicon Valley working on pure vision based next generation self-driving cars. From 2017 to 2020, he was a Principal Researcher at Microsoft Cloud and AI working on real-time 4D computer vision.
Host: Professor Ziming Zhang