Oregon State University | Computer Vision, Deep Learning and Robotics

Dr. Fuxin Li

Professor (effective Sept. 16th, 2026), School of Electrical Engineering and Computer Science, Oregon State University.

I work on computer vision and deep learning, with current emphasis on point cloud deep learning, robotic perception, 3D world models, explainable deep learning and uncertainty estimation.

Point Clouds 3D Perception for Robotics Explainable Deep Learning Uncertainty

Google Scholar X LinkedIn OSU Profile CV

Email fuxin.li@oregonstate.edu

Office 2077 Kelley Engineering Center

Phone (541) 737-5987

Address Corvallis, OR 97331

Research

Deep Machine Vision

My research group Deep Machine Vision (DMV) studies models that see and reason about the 3D real world: from dense visual inputs to sparse 3D observations, reasoning through incomplete geometry and object dynamics, as well as the explainability and uncertainty of the deep network predictions.

We are part of the DRAIL Lab that aims to make humanoid robots truly intelligent.

Point Cloud Deep Learning

PointConv, PointConvFormer, PointPWC-Net and point-based representations for efficient 3D perception.

3D World Modeling

Wide-coverage 3D reconstruction, Gaussian splatting semantic SLAM, spatial-temporal imagination including point cloud completion and 3D dynamic models.

Explainable AI

Develop attribution methods and structured explanations to analyze of how CNNs, transformers and vision-language models make decisions.

Uncertainty and Robustness

Bayesian deep learning, diverse generative models, and methods for estimating when a model should be trusted.

Recent Highlights

What is new

May. 2026

Promotion to full professor.

Fuxin Li has been promoted to Full Professor in the School of Electrical Engineering and Computer Science at Oregon State University, effective September 16, 2026.

May. 2026

Learning a Particle Dynamics Model with Real-world Videos appears in CVPR Findings 2026.

The paper learns particle dynamics directly from real-world videos using rendering supervision in a Gaussian-splatting representation.

Apr. 2026

Softmax-GS appears in CVPR Findings 2026.

The project studies how generalized Gaussians can learn when to blend smoothly and when to preserve sharper boundaries in 3D Gaussian Splatting.

Apr. 2026

Long-LRM++ appears in CVPR Findings 2026.

Long-LRM++ preserves fine details in feed-forward wide-coverage reconstruction and supports scene-level Gaussian reconstruction with real-time rendering.

Mar. 2026

RAMP-3D is accepted to ICAPS 2026.

The work grounds vision and language to 3D masks for long-horizon box rearrangement, linking 3D perception with natural-language robotic planning.

Jan. 2026

OSU feature on computer vision and spatial intelligence.

The story highlights point cloud deep learning, world models from partial observations, Gaussian splatting, and industry-facing robotics applications.

2025

Program Chair, CVPR 2025.

CVPR is the flagship conference in computer vision, with 13k submissions in 2025. We have made 11 reforms to the conference organization. Many of those were adopted in later conferences.

2025

Long-LRM accepted to ICCV 2025 as a highlight.

Long-LRM performs high-resolution, wide-coverage Gaussian reconstruction from long image sequences in a feed-forward model which reconstruct a large scene in 1 second.

2024

CVPR Best Student Paper Runner-Up.

This paper utilizes the deep learning explainability toolbox we built with heatmaps and structural explanations, and uncover two modes of decision-making in visual deep networks: compositional and disjunctive. It also established a significant connection between the choice of normalization layer and the decision-making process of the networks.

People

Deep Machine Vision group

Current Ph.D. Students

Mingqi Jiang
David Smerkous
Pritesh Verma
Wesley Khademi
Skand Peri, co-advised with Stefan Lee
Zhanpei Fang, co-advised with Jamon Van Den Hoek

Other Research Staff

Assistant Professor (Senior Research) Dr. Chanho Kim
M.S. student Bryan Olmstead
Research assistant Yuxin Peng
Undergrad student Bryan Chen

Alumni

Hung Nguyen, Ph.D., Google
Ziwen Chen , Ph.D., Adobe
Jialin Yuan, Ph.D., Meta
Amin Ullah, Postdoc, Boeing
Saeed Khorram, Ph.D., Apple
Wenxuan Wu, Ph.D., Momenta
Robert DeBortoli, Ph.D., Agility Robotics
Neale Ratzlaff, Ph.D., Oracle
Xingyi Li, Ph.D., Samsung
Zhongang Qi, Postdoc, Chief Scientist at Vivo
Rahul Sawhney, Ph.D., Microsoft Research
Venkat Kalyanakumar, M.S., Apple
Tim Player, M.S., Meta
Michael Lowell, M.S.
Mazen Alotaibi, M.S., Malaa Technologies
Ali Behnoudfar, M.S.
Jay Patravali, M.S., Microsoft
Damanpreet Kaur, M.S., Microsoft
Zehuan Chen, M.S., Amazon
Lawrence Neal, M.S., Overland Robotics
Xinyao Wang, M.S., JD Digits, now at Meta
Alrik Firl, M.S., General Electric
Xin Li, M.S., NVIDIA
Zheng Zhou, M.S., Tencent AI Lab

Publications

Selected Recent Publications

I began publishing under the name Li Fuxin in 2019. However, I'm not disciplined enough with all my collaborators so you still see a mix of both Fuxin Li and Li Fuxin in publications. Anyways, Fuxin is my first name, Li is my family name, and you can call me Fuxin when we meet.

2026

CVPR Findings 2026 Softmax-GS: Generalized Gaussians Learning When to Blend or Bound.
Chen Ziwen, Peng Wang, Hao Tan, Zexiang Xu, Li Fuxin.

Project arXiv
CVPR Findings 2026 Long-LRM++: Preserving Fine Details in Feed-Forward Wide-Coverage Reconstruction.
Chen Ziwen, Hao Tan, Peng Wang, Zexiang Xu, Li Fuxin.

Project arXiv
CVPR Findings 2026 Learning a Particle Dynamics Model with Real-world Videos.
Chanho Kim, Suhas V. Sumukh, Li Fuxin.

Project arXiv
ICAPS 2026 Grounding Vision and Language to 3D Masks for Long-Horizon Box Rearrangement.
Ashish Malik, Caleb Lowe, Aayam Shrestha, Stefan Lee, Fuxin Li, Alan Fern.

arXiv

2025

ICCV 2025 Highlight Long-LRM: Long-Sequence Large Reconstruction Model for Wide-Coverage Gaussian Splats.
Ziwen Chen, Hao Tan, Kai Zhang, Sai Bi, Fujun Luan, Yicong Hong, Fuxin Li, Zexiang Xu.

Project arXiv
ICLR 2025 Point-based Instance Completion with Scene Constraints.
Wesley Khademi, Fuxin Li.

OpenReview ICLR poster
AAAI 2025 Data Augmentation Approaches for Satellite Imagery.
Laurel M. Hopkins, Weng-Keen Wong, Hannah Kerner, Fuxin Li, Rebecca A. Hutchinson.

AAAI
arXiv 2025 GS4: Generalizable Sparse Splatting Semantic SLAM.
Mingqi Jiang, Chanho Kim, Chen Ziwen, Li Fuxin.

arXiv Project
arXiv 2025 Where do Large Vision-Language Models Look at when Answering Questions?
Xiaoying Xing, Chia-Wen Kuo, Fuxin Li, Yulei Niu, Fan Chen, Ming Li, Ying Wu, Longyin Wen, Sijie Zhu.

arXiv

2024

CVPR 2024 Best Student Paper Runner-Up Comparing the Decision-Making Mechanisms by Transformers and CNNs via Explanation Methods.
Mingqi Jiang, Saeed Khorram, Li Fuxin.

Paper arXiv
CVPR 2024 Object Dynamics Modeling with Hierarchical Point Cloud-based Representations.
Chanho Kim, Li Fuxin.

CVF arXiv
CVPR 2024 Taming the Tail in Class-Conditional GANs: Knowledge Sharing via Unconditional Training at Lower Resolutions.
Saeed Khorram, Mingqi Jiang, Mohamad Shahbazi, Mohamad H. Danesh, Li Fuxin.

CVF arXiv Code
NeurIPS 2024 Enhancing Diversity in Bayesian Deep Learning via Hyperspherical Energy Minimization of CKA.
David Smerkous, Qinxun Bai, Li Fuxin.

arXiv Code
ICRA 2024 Point Cloud Models Improve Visual Robustness in Robotic Learners.
Skand Peri, Iain Lee, Chanho Kim, Li Fuxin, Tucker Hermans, Stefan Lee.

arXiv

Selected earlier work

CVPR 2023 AutoFocusFormer: Image Segmentation off the Grid.
Ziwen Chen, Kaushik Patnaik, Shuangfei Zhai, Alvin Wan, Zhile Ren, Alex Schwing, Alex Colburn, Li Fuxin.

arXiv Code
CHIL 2021 iGOS++: Integrated Gradient Optimized Saliency by Bilateral Perturbations.
Saeed Khorram*, Tyler Lawson*, Li Fuxin. First two authors contributed equally.

arXiv Code
ICLR 2021 Spotlight Topology-Aware Segmentation Using Discrete Morse Theory.
Xiaoling Hu, Yusu Wang, Li Fuxin, Dimitris Samaras, Chao Chen.
ECCV 2020 Spotlight Point-PWCNet: A Coarse-to-Fine Network for Supervised and Self-Supervised Scene Flow Estimation on 3D Point Clouds.
Wenxuan Wu, Zhiyuan Wang, Zhuwen Li, Wei Liu, Li Fuxin.

arXiv Code
ICLR 2020 Efficient Riemannian Optimization on the Stiefel Manifold via the Cayley Transform.
Jun Li, Li Fuxin, Sinisa Todorovic.

arXiv
NeurIPS 2019 Topology-Preserving Deep Image Segmentation.
Xiaoling Hu, Li Fuxin, Dimitris Samaras, Chao Chen.

arXiv
CVPR 2019 PointConv: Deep Convolutional Networks on 3D Point Clouds.
Wenxuan Wu, Zhongang Qi, Li Fuxin.

arXiv Code
ICML 2019 HyperGAN: A Generative Model for Diverse, Performant Neural Networks.
Neale Ratzlaff, Li Fuxin.

arXiv Code
ECCV 2018 Open Set Learning with Counterfactual Images.
Lawrence Neal, Matthew Olson, Xiaoli Fern, Weng-Keen Wong, Fuxin Li.

Paper Code
ICCV 2015 Oral Multiple Hypothesis Tracking Revisited.
Chanho Kim, Fuxin Li, Arridhana Ciptadi, James M. Rehg.

Paper
ICCV 2013 Video Segmentation by Tracking Many Figure-Ground Segments.
Fuxin Li, Taeyoung Kim, Ahmad Humayun, David Tsai, James M. Rehg.

Project SegTrack v2
CVPR 2013 Composite Statistical Inference for Semantic Segmentation.
Fuxin Li, Joao Carreira, Guy Lebanon, Cristian Sminchisescu.

Paper Technical report
IJCV 2012 Object Recognition as Ranking Holistic Figure-Ground Hypotheses.
Joao Carreira, Fuxin Li, Cristian Sminchisescu. First two authors contributed equally.

Paper

Bio

Background

I am a full professor, effective September 16, 2026, in the School of Electrical Engineering and Computer Science at Oregon State University. My research is broadly in machine learning and computer vision, with a major interest in using and designing learning algorithms for structured data in images, videos, and 3D scenes.

Before joining OSU, I spent several years at Georgia Tech, first as a postdoctoral researcher supervised by Dr. Guy Lebanon and then as a research scientist working with Dr. James M. Rehg. I have also worked on natural language processing, recommender systems, segmentation, object recognition, multi-target tracking, and semantic reconstruction.

Earlier, I was a research scientist in Cristian Sminchisescu's group at the University of Bonn, where I worked on machine learning and computer vision. I was part of the BONN-SVRSEGM team that won the PASCAL VOC semantic segmentation challenges from 2009 to 2012.

I received my bachelor's degree from Zhejiang University and my Ph.D. from the Institute of Automation, Chinese Academy of Sciences, with a dissertation on Euclidean metric learning advised by Jue Wang.

Teaching

Courses taught

CS/AI 535 Deep Learning
CS/AI 637 Computer Vision II
AI 539 Differential and Information Geometry
CS 261 Data Structures