About Me
I am an undergraduate student at Yuanpei College, Peking University (PKU), supervised by Prof. Wenzheng Chen at the Visual Computing and Learning (VCL) Lab. I am also privileged to work closely with Prof. Yixin Zhu at Peking University and Kaichen Zhou at Massachusetts Institute of Technology.
My research centers on closing the Real-Sim-Real loop for robot learning — capturing real-world environments into high-fidelity digital twins, training generalizable policies at scale in simulation, and deploying back to the physical world. I’m always open to research discussions and collaboration. Feel free to reach out via changxinhai@pku.edu.cn or WeChat.
Research Interests
I work on the Real-Sim-Real loop for embodied AI, spanning 3D vision and robot learning.
- 3D Vision: High-fidelity reconstruction, 3D generation, digital twins, and scene understanding.
- Embodied AI: Simulation, world models, mobile manipulation, and sim-to-real transfer.
News
- [Jun. 2026] 🎉 GEM-4D is accepted by ECCV 2026.
- [Apr. 2026] 🏆 AutoMoMa is accepted by CVPR 2026 as Highlight.
- [Mar. 2026] 🎉 EpiS is accepted by CVPR 2026 Workshop A2AMML.
- [Jan. 2026] 🎉 PAGE-4D is accepted by ICLR 2026.
- [Jun. 2025] 🤖 AutoMoMa is accepted by RSS 2025 MoMa.
- [Jun. 2025] 🎤 TerraX is accepted by IROS 2025 as Oral.
- [Feb. 2025] ✨ SplatMesh is accepted by CVPRW 2025.
Preprints
-
Under Review
RAD: A Realistic Multi-View Benchmark for Pose-Agnostic Anomaly Detection
A robot-captured multi-view anomaly benchmark targeting pose variation, reflective materials, and viewpoint-dependent defects.
-
Preprint
Stream3D: Sequential Multi-View 3D Generation via Evidential Memory
A sequential multi-view 3D generation method using evidential memory for consistent high-quality generation.
-
Preprint
NSL-SLAM: High-Fidelity Neural Structured-Light Depth for Practical SLAM and Reconstruction
A high-fidelity neural structured-light depth estimation method for practical SLAM and reconstruction.
Publications
Use Selected for representative works, or switch to All for the full list.
-
ECCV 2026
GEM-4D: Geometry-Enhanced Video World Models for Robot Manipulation
A geometry-enhanced video world model that leverages geometric priors for robot manipulation.
-
CVPR 2026 Highlight
AutoMoMa: Scalable Coordinated Mobile Manipulation Trajectory Generation
A scalable GPU-accelerated trajectory generation system for whole-body mobile manipulation across homes, robots, and tasks.
-
ICLR 2026
PAGE-4D: Disentangled Pose and Geometry Estimation for 4D Perception
A feed-forward 4D perception model that disentangles pose and geometry for dynamic-scene reconstruction.
-
IROS 2025 Oral
TerraX: Visual Terrain Classification Enhanced by Vision-Language Models
A vision-language terrain classification framework combining TerraData, TerraBench, and TerraCLIP for outdoor robotics.
-
CVPR 2026 Workshop
Neural Surface Reconstruction from Sparse Views Using Epipolar Geometry
A sparse-view neural surface reconstruction method that injects epipolar geometry into multi-view feature fusion.
-
CVPRW 2025
SplatMesh: Interactive 3D Segmentation and Editing Using Mesh-Based Gaussian Splatting
An interactive 3D segmentation and editing method that combines Gaussian splatting with mesh-based structure.
-
SMC 2025
AADNet: A Human-Mind-Inspired Multi-Modal Framework for Object Concept Learning
A multi-modal object concept learning framework inspired by human cognitive association and abstraction.
-
ICIC 2025 Oral
StyleCraft: High-Quality Arbitrary Style Transfer via Unified Content-Style Fusion
A unified content-style fusion method for high-quality arbitrary image style transfer.
No publications match this filter yet.
Honors and Awards
- [2025] National Scholarship
- [2024] Second-class Scholarship, Peking University
- [2024] First Prize, 21st Jiang Zehan Cup Mathematical Modeling Competition
- [2023] Xing Zhengde Scholarship, Yuanpei College
- [2023] Third-class Scholarship, Peking University
Experiences
Peking University, China
Undergraduate Student, Data Science and Artificial Intelligence
Visual Computing and Learning (VCL) Lab, Peking University, China
Student Researcher
Advisor: Prof. Wenzheng Chen
Cognitive Reasoning (CoRe) Lab, Peking University, China
Student Researcher
Advisor: Prof. Yixin Zhu
International Asian Liver Center, Gansu, China
Volunteer Teacher
Services
- Teaching Assistant: Introduction to Computation B (Autumn 2025, by Prof. Jun Sun)
- Teaching Assistant: Introduction to Computation C (Autumn 2025, by Prof. Baobao Chang)
- Teaching Assistant: Introduction to Computer System (Autumn 2024, by Prof. Lu Zhang)