Bio

I'm currently a master student at ETH Zurich. I received my Bachelor's degree from Huazhong University of Science and Technology. I work with Dr. Dániel Béla Baráth and Prof. Marc Pollefeys at CVG for my current research. Previously, I worked with Prof. Marco Hutter and Prof. Peidong Liu.

My research interests lie in computer vision and robotics, with a focus on 3D computer vision (e.g. 3D reconstruction, SLAM). I am actively looking for the Ph.D. position in 3D vision and robotics starting from Fall 2026.

News

Feb 2026: DROID-W is accepted to CVPR 2026. See you in Denver!

May 2024: USB-NeRF appears at ICLR 2024.

Publications

* denotes equal contribution or advising; † denotes corresponding author

DROID-SLAM in the Wild

CVPR 2026

Moyang Li, Zihan Zhu, Marc Pollefeys, Dániel Béla Baráth

[Project page] [Paper]

Accurate, robust, and efficient SLAM system under unconstrained, real-world dynamic environments.

USB-NeRF: Unrolling Shutter Bundle Adjusted Neural Radiance Fields

ICLR 2024

Moyang Li*, Peng Wang*, Lingzhe Zhao, Bangyan Liao, Peidong Liu†

[Project page] [Paper] [Code] [Dataset]

USB-NeRF can remove rolling shutter effect, realize high-quality scene reconstruction, and achieve accurate camera trajectory estimation from a sequence of rolling shutter images.

Projects

MDS-NeRF: Neural Radiance Fields with Monocular Depth Supervision

[Project page] [Code]

We utilize affine-invariant depth supervision to address the messy geometry of few-shot scene reconstruction. Our method could estimate the scales and offsets of monocular depth, realize high-quality scene reconstruction with only a few images.

Dynamic Lighting in Neural Radiance Fields

We employ physically-based rendering to decompose the scene into illumination, material, and geometry components. By modelling the illumination with the combination of hemispherical light sources and an environment map, our method can eliminate heavy shadows caused by uneven lighting and improve the quality of scene reconstruction.