Haofei Xu徐豪飞

Researcher at KE:SAI Open Science Lab
Postdoc at ETH Zurich

I am a researcher at KE:SAI Open Science Lab and a postdoc at ETH Zurich, advised by Andreas Geiger and Marc Pollefeys.

I received my PhD from ETH Zurich under the supervision of Marc Pollefeys and Andreas Geiger. During my PhD, I interned at Google, where I worked with Michael Niemeyer and Federico Tombari. Previously, I worked with Jianfei Cai and Hamid Rezatofighi at Monash University. I earned my master's degree from University of Science and Technology of China, where I worked with Juyong Zhang. During my master's, I also studied as an exchange student at NTU and completed a research internship at Microsoft Research Asia.

I am honored to have been named a 2025 Apple Scholar in AI/ML and to have received the Gold Reviewer Award at ICML 2026, the Top Reviewer Award at NeurIPS 2024, and the Outstanding Reviewer Award at CVPR 2022.

Selected Publications

I have broad interests in computer vision and deep learning. I have worked on depth, stereo, optical flow, tracking, feed-forward NeRF/3DGS, and generative models. I am actively exploring new ideas, and I am currently interested in representation learning and generative models for efficient and scalable world models and physical AI. Please see my full publication list on Google Scholar.

PointDiT: Pixel-Space Diffusion for Monocular Geometry Estimation

Haofei Xu, Rundi Wu, Philipp Henzler, Nikolai Kalischek, Michael Oechsle, Fabian Manhardt, Marc Pollefeys, Andreas Geiger, Federico Tombari, Michael Niemeyer

ICML 2026

paper project page code

A minimalist pixel-space diffusion transformer that predicts dense 3D point maps from a single image, with no VAE and no hybrid architecture.

ReSplat: Learning Recurrent Gaussian Splatting

Haofei Xu, Daniel Barath, Andreas Geiger, Marc Pollefeys

ECCV 2026Spotlight

paper slides project page code

Test-time scaling for feed-forward Gaussian splatting.

MegaFlow: Zero-Shot Large Displacement Optical Flow

Dingxi Zhang, Fangjinhua Wang, Marc Pollefeys, Haofei Xu

ECCV 2026

paper project page code demo

A simple yet powerful framework for large-displacement optical flow and point tracking.

Multi-View 3D Point Tracking

Frano Rajič, Haofei Xu, Marko Mihajlovic, Siyuan Li, Irem Demir, Emircan Gündoğdu, Lei Ke, Sergey Prokudin, Marc Pollefeys, Siyu Tang

ICCV 2025Oral

paper project page interactive results code

The first data-driven multi-view 3D point tracker for tracking arbitrary 3D points across multiple cameras.

DepthSplat: Connecting Gaussian Splatting and Depth

Haofei Xu, Songyou Peng, Fangjinhua Wang, Hermann Blum, Daniel Barath, Andreas Geiger, Marc Pollefeys

CVPR 2025

paper project page code

Cross-task interactions between feed-forward Gaussian splatting and depth.

PanSplat: 4K Panorama Synthesis with Feed-Forward Gaussian Splatting

Cheng Zhang, Haofei Xu, Qianyi Wu, Camilo Cruz Gambardella, Dinh Phung, Jianfei Cai

CVPR 2025

paper project page code interactive 4K 360 demo

4K panorama synthesis with a single feed-forward inference.

No Pose, No Problem: Surprisingly Simple 3D Gaussian Splats from Sparse Unposed Images

Botao Ye, Sifei Liu, Haofei Xu, Xueting Li, Marc Pollefeys, Ming-Hsuan Yang, Songyou Peng

ICLR 2025Oral

paper project page code

Unposed 3DGS reconstruction made easy.

MVSplat: Efficient 3D Gaussian Splatting from Sparse Multi-View Images

Yuedong Chen, Haofei Xu, Chuanxia Zheng, Bohan Zhuang, Marc Pollefeys, Andreas Geiger, Tat-Jen Cham, Jianfei Cai

ECCV 2024Oral

paper project page code

A cost volume representation for efficiently predicting 3D Gaussians from sparse multi-view images in a single feed-forward inference.

Unifying Flow, Stereo and Depth Estimation

Haofei Xu, Jing Zhang, Jianfei Cai, Hamid Rezatofighi, Fisher Yu, Dacheng Tao, Andreas Geiger

TPAMI 2023

paper project page slides video(cn) colab demo code

A unified dense correspondence matching formulation enables three motion and 3D perception tasks to be solved with a unified model.

GMFlow: Learning Optical Flow via Global Matching

Haofei Xu, Jing Zhang, Jianfei Cai, Hamid Rezatofighi, Dacheng Tao

CVPR 2022Oral

paper slides video(cn) poster code

Learning cross-view features with a Transformer enables optical flow to be solved by directly comparing feature similarities.

High-Resolution Optical Flow from 1D Attention and Correlation

Haofei Xu, Jiaolong Yang, Jianfei Cai, Juyong Zhang, Xin Tong

ICCV 2021Oral

paper code

Factorizing 2D optical flow with 1D attention and 1D correlation enables 4K resolution optical flow estimation on standard GPUs.

AANet: Adaptive Aggregation Network for Efficient Stereo Matching

Haofei Xu, Juyong Zhang

CVPR 2020

paper code

A sparse points-based cost aggregation method leads to an efficient and accurate stereo matching architecture without any 3D convolutions.

Open Source Projects

Key open-source projects developed through my research and collaborations.

PointDiT

Pixel-Space Diffusion for Monocular Geometry Estimation

ReSplat

Learning Recurrent Gaussian Splatting

MegaFlow

Zero-Shot Large Displacement Optical Flow

MVTracker

Multi-View 3D Point Tracking

NoPoSplat

Surprisingly Simple 3D Gaussian Splats from Sparse Unposed Images

DepthSplat

Connecting Gaussian Splatting and Depth

MVSplat

Efficient 3D Gaussian Splatting from Sparse Multi-View Images

UniMatch

Unifying Flow, Stereo and Depth Estimation

GMFlow

Learning Optical Flow via Global Matching

AANet

Adaptive Aggregation Network for Efficient Stereo Matching

Invited Talks

Generative Reconstruction: Towards Unified Spatial IntelligenceMicrosoft Research Asia (MSRA) · hosted by Jiaolong Yang

Jul 16, 2026

PointDiT: Pixel-Space Diffusion for 3D GenerationZhejiang University · hosted by Yiyi Liao

Apr 23, 2026

PointDiT: Pixel-Space Diffusion for 3D GenerationWestlake University · hosted by Anpei Chen

Apr 22, 2026

Learning to Optimize 3D Gaussian SplattingHuawei

Mar 9, 2026

Multi-View 3D Point TrackingGoogle DeepMind · hosted by Junhwa Hur

Sep 29, 2025

Learning to SplatHuawei

Jun 3, 2025

DepthSplat: Connecting Gaussian Splatting and DepthGoogle DeepMind · hosted by Ben Poole

Oct 29, 2024

Unifying Flow, Stereo and Depth EstimationSynced (机器之心) · slides

Dec 28, 2022

GMFlow: Learning Optical Flow via Global MatchingMonash University · slides

Apr 13, 2022

Teaching

Head Teaching Assistant3D Vision

Spring 2026

Head Teaching Assistant3D Vision

Spring 2025

Teaching AssistantComputer Vision

Fall 2024

Teaching AssistantStochastics and Machine Learning

Spring 2024

Teaching AssistantRobot Learning

Fall 2023

Academic Services

Conference Reviewer

ICCV 2021CVPR 2022ECCV 2022CVPR 2023NeurIPS 2023CVPR 2024ECCV 2024NeurIPS 2024CVPR 2025ICCV 2025NeurIPS 2025CVPR 2026ICML 2026ECCV 2026

Journal Reviewer

TIPIJCVTPAMI

Awards

Gold Reviewer

ICML 2026

Apple Scholar in AI/ML

2025

Top Reviewer

NeurIPS 2024

Outstanding Reviewer

CVPR 2022

1st place of Argoverse Stereo Challenge, CVPR 2022 Workshop on Autonomous Driving

2022

National Scholarship

2016