Towards Efficient 3D Geometric Scene Understanding with Humans

Efficiently understanding the 3D geometry of a scene from visual data, such as images and videos, is critical for real-world applications. Intelligent agents, such as robots or self-driving cars, use this information to navigate and interact with the world. Moreover, as humans are a critical dynamic factor in the environment, understanding and predicting human motion will benefit the human-intelligent agent interaction. In this talk, I will first cover our past and ongoing work in efficient depth map estimation, deterministic and stochastic 3D human motion predictions, and 3D scene-aware human motion prediction and generation. Then, I will provide insights on how our research benefits education and, finally, our collaborations with the industry.

Join Zoom Meeting

Meeting ID: 897 0608 6809

Password: 717466


Dr. Miaomiao Liu


Dr. Miaomiao Liu is a Senior Lecturer and an ARC DECRA Fellow in the School of Computing, the Australian National University. She was a Research Scientist in Data 61, CSIRO (2017-2018), and a researcher in NICTA (2012-2016). She earned her Ph.D. (2012) from the University of Hong Kong, China. Her research interest is mainly in Computer Vision, 3D vision, 3D geometric scene understanding, and understanding human 3D body shape and motions. She currently serves as a Senior PC for IJCAI 2022 and an Associate Editor for IEEE Robotics and Automation Letters (RA-L) (2021 -).

Date & time

10–11am 22 Oct 2021



Updated:  10 August 2021/Responsible Officer:  Dean, CECS/Page Contact:  CECS Marketing