Biography
Hi, I am Yun He, a third-year CS PhD student at the University of Maryland, College Park, advised by Prof. Matthias Zwicker. I received my master's degree in computer science from Fudan University in 2023, co-advised by Prof. Yanwei Fu and Prof. Xiangyang Xue. During my master's studies, I also collaborated closely with Dr. Danhang Tang and Dr. Yinda Zhang at Google. In 2020, I obtained my bachelor's degree in computer science from Chongqing University.
My primary research interests include:
- Video Generation and Editing, particularly for autonomous vehicles and robotics, such as LangDriveCTRL (ECCV 2026).
- Multimodal Agents, with an emphasis on building efficient and automated agentic frameworks, ranging from a training-free framework (LangDriveCTRL, ECCV 2026) to trainable orchestration (SceneOrchestra, ECCV 2026).
- 3D Vision, including Gaussian Splatting for object relighting (TranSplat, ICCP 2026), as well as point cloud processing for upsampling (Grad-PU, CVPR 2023) and compression (D-PCC, CVPR 2022).
Interests
- Video Generation and Editing
- Multimodal Agents
- 3D Vision
Education
-
Ph.D. in Computer Science, 2024-Present
University of Maryland, College Park
-
M.Sc. in Computer Science, 2020-2023
Fudan University
-
B.Eng. in Computer Science, 2016-2020
Chongqing University
News
- [2026-06] LangDriveCTRL and SceneOrchestra are accepted to ECCV 2026.
- [2026-06] I join Amazon as an Applied Scientist Intern.
- [2026-05] TranSplat is accepted to ICCP 2026.
- [2026-04] Parallel-Probe is accepted to ICML 2026.
- [2025-12] We release LangDriveCTRL, an agentic workflow for driving video editing.
- [2025-05] I join NEC Labs America as a Research Intern.
- [2024-08] I join UMD as a CS PhD student.
- [2023-06] I am awarded as an Outstanding Master's Student in Shanghai (Top 3%)!
- [2023-02] Grad-PU is accepted by CVPR 2023. Code is avaliable at Github.
- [2022-02] D-PCC is accepted by CVPR 2022. Code is avaliable at Github.
Publications
Yun He, Kelin Yu, Matthias Zwicker
Yun He, Francesco Pittaluga, Ziyu Jiang, Matthias Zwicker, Manmohan Chandraker, Zaid Tasneem
Boyang Yu, Yanlin Jin, Yun He, Akshat Dave, Ravi Ramamoorthi, Guha Balakrishnan
Yun He, Danhang Tang, Yinda Zhang, Xiangyang Xue, Yanwei Fu
Yun He*, Xinlin Ren*, Danhang Tang, Yinda Zhang, Xiangyang Xue, Yanwei Fu
Tong Zheng, Chengsong Huang, Runpeng Dai, Yun He, Rui Liu, Xin Ni, Huiwen Bao, Kaishen Wang, Hongtu Zhu, Jiaxin Huang, Furong Huang, Heng Huang
Experience
Applied Scientist Intern
- Develop an arrow-guided method for controllable driving video generation: users select a target vehicle and specify its motion with an arrow in the first frame, then the generated video shows the vehicle following that motion.
- Achieve strong motion controllability and high visual quality in the generated videos.
Research Intern
- Develop LangDriveCTRL (ECCV 2026), an agentic framework for driving video editing that supports both object-level editing (removal, insertion, and replacement) and multi-object behavior editing.
- Achieve both high controllability and strong realism by combining explicit modular agents with diffusion-based refinement.
Honors and Awards
Dean’s Fellowship
Outstanding Master’s Student Award in Shanghai (Top 3%)
Outstanding Master’s Thesis Award at Fudan University (Top 10%)
Outstanding Teaching Assistant Award at Fudan University (Top 10%)