Email  /  Scholar  /  LinkedIn  /  X  /  Github

Maolin Gao (高茂林)

I am a Ph.D. student at TUM CVAI Lab with Prof. Daniel Cremers, working on computer vision, graphics and optimisation. I was a visiting scholar at the Geometric Computing Lab at Stanford University headed by Prof. Leonidas J. Guibas in the summer 2024.

Prior to my PhD, I worked as a research scientist at Artisense, a SLAM startup co-founded by Prof. Daniel Cremers, which has been successfully acquired by Kudan and a computer vision engineer on ADAS Vision systems at Bosch Stuttgart. Prior to this, I obtained my Master's degree at TUM Department of Electrical Engineer and Information Technology, where I wrote my master thesis on Blind Deconvolution under the supervision of Dr. Michael Hirsch, Prof. Philipp Hennig and Prof. Bernhard Schölkopf in MPI-IS Tübingen. I obtained my Bachelor in Physics at Tongji University in Shanghai, during which I spent a year at TUM and a summer at RWTH Aachen.

• 2026/04: Our paper on SO(3)-Equivariant and invariant manifold learning has been accepted in CVPR 2026 as a highlight, project page.
• 2026/04: Our paper on a fast batch functional map solver got accepted in CVPR Image Matching Workshop 2026 in Denver, USA.
• 2025/06: Our partial shape dataset and benchmark paper has been accepted in SGP 2025 in Bilbao, Spain.
• 2025/03: Our paper on partial-to-partial shape matching called echoMatch has been accepted in CVPR 2025 in Nashville, USA.
• 2024/11: Our paper on embedding learning for point clouds got accepted in 3DV 2025 in Singapore.
• 2024/08: One paper on neural surface reconstruction will apear in GCPR 2024 in Munich, Germany.
• 2024/07: I will visit the Geometric Computing Lab at Stanford and working with Prof. Leonidas J. Guibas for this summer.
• 2024/03: Two papers have been accepted in CVPR 2024 in Seattle, USA.
      1. Finsler-Laplace-Beltrami Operators with Application to Shape Analysis.
      2. Partial-to-Partial Shape Matching with Geometric Consistency.
• 2023/10: Our paper on consistent partial shape matching has been accepted in 3DV 2024 in Davos, Switzerland.
• 2023/08: Our paper on global shape matching has been accepted in ICCV 2023 in Paris, France.
• 2021/03: Our paper on isometric multi-shape matching has been accepted as oral in CVPR 2021.
• 2020/10: Our paper on distributed photometric BA has been accepted as oral in 3DV in Fukuoka, Japan.
• 2020/01: Moved back to CVAI Lab at TUM.
• 2020/01: Artisense.ai has been acquired by Kudan.
• 2019/05: Our paper on general uncalibrated PS has been accepted in ICCV 2019 in Seoul, Korea.
• 2018/11: Joined Artisense.ai, an autonomous driving startup in Munich.
• 2016/01: Joined Bosch ADAS vision department in Stuttgart.

A rotation-invariant manifold/surface learning architecture for understanding 3D shapes.

A framework to generate partial shapes with ground truth correspondences across datasets.

Cycle-consistency applied to partial shapes for accurate overlapping region prediction.

A geometrically motivated shape embedding that is robust to non-rigid deformations.

A solution of heat equation on the Finsler manifold which induces natural anisotropy.

Partial-to-partial shape matching approaches guided by geometric consistency.

Partial-to-full shape matching approaches encourages geometric consistency.

A scale and E(3)-invariant shape deformation operator and a matching method which can be solved to global optimality.

An efficient and easy-to-implement isometric multi-shape matching algorithm utilising a universe shape and functional maps.

An approximate consensus technique to optimise large-scale bundle adjustment in distribution.

A variational approach to enable photometric stereo outside of laboratory conditions.

Web template © this.   All content © Maolin Gao