Hongrui Cai (蔡泓锐)

I am a Ph.D. candidate in the Graphics&Geometric Computing Laboratory (GCL) at University of Science and Technology of China (USTC), supervised by Prof. Juyong Zhang. Before that, I obtained my B.S. degree from South China University of Technology (SCUT) in 2019.

Email: hrcai AT mail.ustc.edu.cn

I am looking for formal job opportunities in 2024!

Google Scholar  /  Github (1.8k+ star)  /  Twitter  /  LinkedIn

CV (English)  /  CV (Chinese)

profile photo
Research Interests

My research interests include some sub-fields of Computer Vision and Graphics:

  • Avatar Modeling and Animation

  • 3D Reconstruction and Structure-from-motion

  • Point Cloud Learning and Surface Registration

CIMS2023 Differentiable Deformation Graph based Neural Non-rigid Registration
Wanquan Feng, Hongrui Cai, Junhui Hou, Bailin Deng, Juyong Zhang.

Communications in Mathematics and Statistics (CIMS), 2023
paper / video

We propose a differentiable deformation graph based neural learning method for non-rigid registration by replacing some components with neural based strategies to fully take advantage of the shape priors and domain knowledge embedded in trained neural networks.


Neural Surface Reconstruction of Dynamic Scenes with Monocular RGB-D Camera
Hongrui Cai, Wanquan Feng, Xuetao Feng, Yan Wang, Juyong Zhang.

NeurIPS, 2022 (Spotlight)
project page / paper / OpenReview / poster / code

We propose Neural-DynamicReconstruction (NDR), a template-free method to recover high-fidelity geometry and motions of a dynamic scene from a monocular RGB-D camera. To represent and constrain the non-rigid deformations, we propose a novel neural invertible deforming network such that the cycle consistency between arbitrary two frames is automatically satisfied.


CariPainter: Sketch Guided Interactive Caricature Generation
Xin Huang, Dong Liang, Hongrui Cai, Juyong Zhang, Jinyuan Jia.

ACM MM, 2022
paper & video

We propose CariPainter, the first interactive caricature generating and editing method, by utilizing the semantic segmentation maps as an intermediary domain.


Neural Points: Point Cloud Representation with Neural Fields for Arbitrary Upsampling
Wanquan Feng, Jin Li, Hongrui Cai, Xiaonan Luo, Juyong Zhang.

CVPR, 2022
project page / paper / zhihu / code

In this paper, we propose Neural Points, a novel point cloud representation. Different from traditional point cloud representation, each point in Neural Points represents a local continuous geometric shape via neural fields. Therefore, Neural Points can express more complex geometry shapes.


Recurrent Multi-view Alignment Network for Unsupervised Surface Registration
Wanquan Feng, Juyong Zhang, Hongrui Cai, Haofei Xu, Junhui Hou, Hujun Bao.

CVPR, 2021
project page / paper / zhihu / code

For non-rigid registration, we propose RMA-Net to deform the input surface shape stage by stage. RMA-Net is trained in an unsupervised manner via our proposed multi-view 2D projection loss.


Landmark Detection and 3D Face Reconstruction for Caricature using a Nonlinear Parametric Model
Hongrui Cai, Yudong Guo, Zhuang Peng, Juyong Zhang.

Graphical Models (GMOD), 2021
paper / slides / zhihu / code

To the best of our knowledge, this is the first work for automatic landmark detection and 3D face reconstruction for general caricatures.


Real-time Face View Correction for Front-facing Cameras
Yudong Guo, Juyong Zhang, Yihua Chen, Hongrui Cai, Zhangjin Huang, Bailin Deng.

Computational Visual Media (CVM), 2021
paper / video

We take the video stream from a single RGB camera as input, and generate a video stream that emulates the view from a virtual camera at a designated location.

submission Double References Guided Interactive 2D and 3D Caricature Generation
Xin Huang, Dong Liang, Hongrui Cai, Yunfeng Bai, Juyong Zhang, Jinyuan Jia.


We propose the first geometry and texture (double) referenced interactive 2D and 3D caricature generating and editing method.

submission Self-Supervised Topology-Aware Non-Rigid Point Cloud Registration Zhuang Peng*, Hongrui Cai*, Juyong Zhang (*equal contribution).


We propose a topology-aware module to perceive topological changes in motion to improve the performance of the algorithm in the dynamic topology region.

Technical Projects
project Landmark driven Facial Expression Recognition
Hongrui Cai.


A landmark-driven Facial Expression Recognition (FER) method without employing any pre-trained models from other tasks

Ant Group
May. 2023 - Present
Research Intern @ Interaction Intelligence Lab
University of Science and Technology of China
Sep. 2019 - Present
Ph.D. candidate in Data Science
South China University of Technology
Sep. 2015 - Jun. 2019
Bachelor Degree in Mathematics and Applied Mathematics
Professional Activities

Conference Reviewers of CVPR 2023, ICCV 2023.

Journal Reviewers of IEEE Transactions on Multimedia (TMM), Computers & Graphics (C&G).

Academic Talks

[2022.12] Invited talk about NDR at NeurIPS 2022.

[2022.11] Invited talk about NDR at Jiqizhixin.

[2021.04] Invited talk about Caricature Face at CVM 2021.

Selected Honors

First-class Academic Scholarships for Postgraduates, by USTC, 2019 - 2023.

Excellent Undergraduate Thesis Award, by SCUT, 2019.

Excellent Undergraduate Student, by SCUT, 2019.

Last update: 2023.11