Miaomiao Liu

Miaomiao Liu is a computer vision researcher at the Australian National University whose work covers 3D scene reconstruction and 3D human motion prediction. She earned her PhD at the University of Hong Kong in 2012 under Kwan-Yee Kenneth Wong, then worked with Richard Hartley as a postdoctoral researcher, at NICTA as a researcher, and at Data61, CSIRO as a research scientist before starting to lead an ANU research project in 2018. At ICCV 2019 and ECCV 2020 she co-authored human motion prediction papers with Wei Mao, and CVPR carried her work on multi-view stereo in 2020, action-driven motion prediction in 2022, online scene reconstruction in 2023 and neural rendering in 2024. She is an Associate Professor in the School of Computing at ANU, where she convenes MMLCV and leads a lab working on 3D vision, human motion understanding and acoustic field learning.

Now at: Australian National University

Timeline

2012

A PhD at Hong Kong

In November 2012 she earns her PhD from the Department of Computer Science at the University of Hong Kong, supervised by Kwan-Yee Kenneth Wong.
2012–2016

Postdoc with Richard Hartley

From 2012 to 2014 she is a postdoctoral researcher working with Richard Hartley. In 2014 she moves to NICTA, the National ICT Centre of Australia, as a researcher.
2016–2018

Research scientist at Data61

In Canberra, she works as a research scientist at Data61, CSIRO, from 2016 to 2018, on 3D vision and 3D scene modeling and understanding.
2018

Machine vision for solar power

In April 2018 she becomes principal investigator of the ANU project "Machine Vision Techniques for Solar Power Forecasting and Generation", which runs until July 2023. At ANU she is an ARC DECRA fellow before becoming an associate professor.
2019–2020

Human motion prediction papers

With Wei Mao, Mathieu Salzmann and Hongdong Li she co-authors "Learning Trajectory Dependencies for Human Motion Prediction", an oral at ICCV 2019 that proposes a feed-forward network for motion prediction. In 2020 she co-authors "History Repeats Itself: Human Motion Prediction via Motion Attention" with Wei Mao and Mathieu Salzmann for ECCV 2020; its attention-based feed-forward network exploits the way human motion tends to repeat itself.
2020

Cost volume pyramid network

She co-authors "Cost Volume Pyramid Based Depth Inference for Multi-View Stereo" with Jiayu Yang, Wei Mao and Jose M. Alvarez, a CVPR 2020 oral. The model infers depth from multi-view images and runs 6x faster with performance similar to state-of-the-art methods.
2022–2024

Motion, VisFusion and HashPoint

In 2022 she co-authors a CVPR 2022 oral with Wei Mao and Mathieu Salzmann that introduces action-driven stochastic human motion prediction from a sequence of action labels. In 2023 she co-authors VisFusion with Huiyu Gao and Wei Mao, a CVPR 2023 method for visibility-aware online 3D scene reconstruction from posed monocular videos. CVPR 2024 carries her HashPoint paper on accelerated point searching and sampling for neural rendering.
As of October 2026

Associate professor at ANU

She is an Associate Professor in the School of Computing at the Australian National University and MMLCV convener. In 2025 her lab of six students works on 3D vision, geometric scene understanding, human motion understanding and acoustic field learning. She is principal investigator of the ANU project "Multi-modal large-scale 3D environment monitoring and conservation", which runs from November 2022 to February 2027.