About me

I am currently a research assistant in Electronic Information School, Wuhan University, China, supervised by Prof. Jiayi Ma. I am an incoming Ph.D. student of École Polytechnique Fédérale de Lausanne (EPFL), Switzerland. Before that, I obtained the MEng. degree and the BEng. degree from the same university. My research is mainly related to 3D computer vision and geometry processing. I am also broadly interested in computer graphics and computer vision in general. Currently I am working on 3D deformable shape matching and 3D geometry learning with deep learning and NeRF techniques.

You can find my full CV here.


Jump to: 2022, 2021, 2020.


Coherent Point Drift Revisited for Non-Rigid Shape Matching and Registration
Aoxiang Fan, Jiayi Ma, Xin Tian, Xiaoguang Mei, Wei Liu
Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022
paper code


Efficient Deterministic Search with Robust Loss Functions for Geometric Model Fitting
Aoxiang Fan, Jiayi Ma, Xingyu Jiang, Haibin Ling
IEEE Transactions on Pattern Analysis and Machine Intelligence
paper code

Smoothness-Driven Consensus Based on Compact Representation for Robust Feature Matching
Aoxiang Fan, Xingyu Jiang, Yong Ma, Xiaoguang Mei, Jiayi Ma
IEEE Transactions on Neural Networks and Learning Systems
paper code

Image matching from handcrafted to deep features: A survey
Jiayi Ma, Xingyu JiangAoxiang Fan, Junjun Jiang, Junchi Yan
International Journal of Computer Vision (IJCV), 2021


Geometric Estimation via Robust Subspace Recovery
Aoxiang Fan, Xingyu Jiang, Yang Wang, Junjun Jiang, Jiayi Ma
Proc. European Conference on Computer Vision (ECCV), 2020
paper code


Ph.D. in Computer Science
École Polytechnique Fédérale de Lausanne (EPFL)
2022 - present

M.Eng. in Information and Communication Engineering
Wuhan University, MVPLab
2018 - 2021

B.Eng. in Electronic Information Science and Technology
Wuhan University
2014 - 2018

Research Experience

TuSimple-Autonomous Trucking Technology
Worked closely with Dr. Ji Zhao and Dr. Naiyan Wang
November 2020 - March 2021

Research Projects

Deep Learning of Feature Matching in the Perspective of Graph Matching

In the field of image feature matching, the emergence of the SuperGlue method which uses a trained network in place of plain nearest neighbor matching, has significantly improved the capacity of many practical applications such as SfM and visual localization. However, the network of SuperGlue essentially condiers a linear assignment problem in its matching process. In this project, we intend to incorporate a graph matching (quadratic assignment) perspective to design a novel network for the task of image feature matching.


Email: aoxiangfan@gmail.com