profile photo

He ZHU | 祝贺

Hi there, thanks for dropping by!

I'm a senior undergrad in Computer Science at Beijing Normal University, and a research intern at Shanghai AI Laboratory, working with Dr. Liang Pan and Prof. Ziwei Liu. Previously, I had the privilege to work with Prof. Xiaowei Zhou and Prof. Sida Peng at the State Key Laboratory of CAD&CG, Zhejiang University.

My passion lies in the intersection of Computer Vision, Computer Graphics and Robotics, with my undergrad research centered on the perception, reconstruction and generation of 3D scenes.

I will join Carnegie Mellon University as a Master's in Robotics student in Fall 2025.

Email  /  Scholar  /  Github

News


  - 02/2025      One paper accepted by CVPR 2025!
  - 07/2024      One paper accepted by ECCV 2024!


Education


B.S. in Computer Science
School of Artificial Intelligence, Beijing Normal University, Beijing, China
GPA: 90.53/3.84; Major GPA: 94.73/4.00
(Majoring in Biology from Sep. 2019 to Jul. 2021)

Sep. 2019 - Jul. 2025 (Expected)

Experience


AIGC Group, Shanghai AI Laboratory, Shanghai, China
Research Intern, Advisor: Dr. Liang Pan & Prof. Ziwei Liu.

Dec. 2024 - Present

State Key Laboratory of CAD&CG, Zhejiang University, Hangzhou, China
Visiting Student, Advisor: Prof. Xiaowei Zhou & Prof. Sida Peng

Aug. 2023 - Nov. 2024

Publication


* denotes equal contribution

Murre: Multi-view Reconstruction via SfM-guided Monocular Depth Estimation

Haoyu Guo*, He Zhu*, Sida Peng, Haotong Lin, Yunzhi Yan, Tao Xie, Wenguan Wang, Hujun Bao, Xiaowei Zhou

CVPR, 2025

project page / paper / code

TL;DR: We propose a sparse-to-dense paradigm for multi-view geometry reconstruction by combining scene structure captured by SfM and monocular depth prior distilled from 2D foundation models.

SAM-guided Graph Cut for 3D Instance Segmentation

Haoyu Guo*, He Zhu*, Sida Peng, Yuang Wang, Yujun Shen, Ruizhen Hu, Xiaowei Zhou

ECCV, 2024

project page / paper / code

TL;DR: We achieve accurate zero-shot instance segmentation on arbitrary 3D scenes by cutting superpoint graphs into objects with SAM-produced features and 2D segmentation masks.






Last update 2025-03-04