Qinfeng Zhu · Research portfolio
From vision
to embodied
intelligence.
I am a visiting researcher in embodied intelligence at Duke University (based Kunshan), supervised by Prof. Kaizhu Huang. From visual understanding to physical action, my current work focuses on agentic robotics safety and humanoid robot collaboration.
Ph.D. student at the University of Liverpool and Xi'an Jiaotong-Liverpool University, advised by Dr. Lei Fan and Dr. Anh Nguyen. MRes in Computer Science, University of Liverpool, 2023.
Feel free to get in touch.
Visiting research · Embodied intelligence
Duke University (based Kunshan). — presentOpen to internship opportunities.
After graduation
Seeking full-time research roles in academia or industry.
01 / Agentic robotics
Choose an object. Watch the arm reach, grasp and place it.
Click an object or run a grasp · drag to orbit
Illustrative kinematics · not a learned policyOpen explorer ↗
Current research
From seeing the world to acting responsibly in it. My latest work studies safety at the boundary between agentic reasoning and physical execution.
A Lease-Enforced Agentic Safety Harness with task, action and commit gates: one shared safety policy, checked from the initial instruction to the moment an action reaches the robot.
Evaluated with RoboRisk: 105 tasks spanning 92 simulation tasks and 13 real-robot tasks.
In progress
Ongoing research based on the Unitree G1. Technical details are not yet public.
Research trajectory
My earlier work asks how machines understand a scene. My current direction asks how that understanding can support reasoning, planning and useful physical action.
Explore the research trajectory ↗Connecting perception, reasoning and action through agents that interact with the physical world.
Exploring world–action models and vision–language–action systems for grounded decisions.
Studying how embodied intelligence can translate into coordinated, precise hand–object interaction.
The foundation · computer vision
My published research in 3D, panoramic and multispectral perception provides the visual and geometric foundation for this next chapter.
Selected work
Selected work in computer vision: the perception and representation foundations of my move toward embodied intelligence.
Search all publications ↗
Complete record
9 further papers · 18 publications across this page
Yuan Fang, Yuanzhi Cai, Jagannath Aryal, Qinfeng Zhu, Hong Huang, Cheng Zhang, Lei Fan
Qinfeng Zhu, Han Li, Liang He, Lei Fan
Qinfeng Zhu, Ningxin Weng, Lei Fan, Yuanzhi Cai
Qinfeng Zhu, Yuan Fang, Lei Fan
Qinfeng Zhu, Yuanzhi Cai, Yuan Fang, Yihan Yang, Cheng Chen, Lei Fan, Anh Nguyen
Qinfeng Zhu, Jiaze Cao, Yuanzhi Cai, Lei Fan
Qinfeng Zhu, Yuanzhi Cai, Lei Fan
Yang Zhao, Dufei Wang, Qinfeng Zhu, Lei Fan, Yuanfeng Bao
All 18 publications are visible on this page, including LEASH in the robotics spotlight above. The library adds search, filters, direct links and one-click BibTeX copying.
Open searchable library ↗Research infrastructure
Curated benchmarks that bridge spectral sensing, dense annotation and practical scene understanding.
View all projects ↗Explore benchmark ↗
Pedestrian tracking from fixed, quadruped and drone viewpoints, paired with consented real panoramas and the Sextant localization readout.
Explore dataset ↗
A large-scale, high-resolution terrestrial laser-scanning dataset for accurate urban scene understanding.
Explore dataset ↗
A dedicated multispectral benchmark for semantic segmentation across diverse indoor scenes.
Latest signals
Recent publication milestones and dataset releases. The complete timeline lives in the Research Atlas.
View timeline ↗LEASH: Stay Safe in Agentic Robotics is now public, with the paper, RoboRisk benchmark and real-robot demonstrations. Under review.
The PanoPed project page and Sextant code are now public, with benchmark formats, license terms and data-access links.
SemanticUrban has been accepted by Expert Systems with Applications.
SwinMamba has been accepted by Digital Signal Processing.
ClassWise-CRF has been accepted by Neural Networks 2026.
IndoorMS has been accepted by IEEE Sensors Journal.
Academic community
Contributing to rigorous and constructive evaluation across computer vision, machine learning and remote sensing.
Membership
Journal reviewer
Conference & workshop reviewer
International Conference on Learning Representations · AAAI Conference on Artificial Intelligence · International Joint Conference on Artificial Intelligence · Machine Learning for Health · IEEE/CVF Winter Conference on Applications of Computer Vision · International Joint Conference on Neural Networks · IEEE International Symposium on Biomedical Imaging · IEEE Symposium on Computers & Informatics · International Conference on Artificial Intelligence and Human-Computer Interaction
Have an idea?
I welcome collaborations in embodied intelligence, agentic robotics and visual perception, as well as research internship opportunities. Feel free to get in touch.
Preferred emailzhuqinfeng1999@gmail.com
