N-CRiPT Public Seminar: Human 3D Pose Estimation From Videos: Challenges and Solutions
Jointly presented with SGInnovate
Speaker: Associate Professor Robby T. Tan
Date & Time: Wednesday 12 May, 11.00am – 12.00pm
With the ubiquitous presence of cameras (e.g. CCTV cameras, mobile phones, etc.), there has been an explosion of images and video data. While there are advantages and convenience of these visual data, they come with cost: potential breach in privacy and confidentiality, particularly for information related to human poses and actions. In this talk, we will focus on discussing the technical challenges and the possible solutions of predicting human 3D poses from a monocular video.
We will discuss some existing top-down and bottom-up approaches of human 3D pose estimation, and explain the problems when occlusions happen or when the ground-truth 3D data is insufficient to capture all possible poses. We then proceed with our solutions for these two important problems. We hope that by the end of the talk, we can know what the state of the art technologies that can be used to analyze the privacy information in videos related to human actions and poses.
A recording of the webinar is available here.
Robby T. Tan is Associate Professor at Yale-NUS College and Department of Electrical and Computing Engineering, National University of Singapore. Previously, he was faculty at Utrecht University, a research associate at Imperial College London, and a research scientist at NICTA/Australian National University. His research interests include computer vision, deep learning and machine learning, particularly in the domains of bad weather, low-level vision, human motion analysis, and fundamental problems of deep learning. He received his PhD degree in computer science from the University of Tokyo.