Hyunjin Kim

I am a MS student in Computer Science at UC San Diego, where I am advised by Prof. Xiaolong Wang. I have also been fortunate to have worked with Prof. Henrik Christensen.

Previously, I received my BS in Urban Planning and Engineering and Computer Science and Engineering from Yonsei University. During my undergraduate studies, I was advised by Prof. Jaesik Park at Seoul National University and was an exchange student at the University of Toronto, where I worked with Prof. Florian Shkurti.

Email:


Actively Seeking PhD Positions for Fall 2026!

How to pronounce my name

My first name is pronounced "He-un-jin," with the "He-un" part spoken quickly. I'm also known as my nickname, HJ. My name in Korean is 김현진.

Google Scholar /  Linkedin /  Github /  CV

profile photo

News

[Sep '25]  

One paper on 3DGS sampling is accepted to NeurIPS 2025!

My research interests encompasses 3D computer vision and robotics. My research goal is to build computational systems capable of human-like scene understanding, such as perceiving complex environments and developing efficient world models that capture visual appearance and physical dynamics. To this extent, my current research focuses on developing physical multimodal world models and efficient 3D/4D scene reconstruction methods.


Metropolis-Hastings Sampling for 3D Gaussian Reconstruction
Hyunjin Kim, Haebeom Jung, Jaesik Park
Neural Information Processing Systems (NeurIPS), 2025
Paper / Website / Code

We introduce a reformulation of 3D Gaussian Splatting (3DGS) as a Metropolis-Hastings sampling process, replacing its heuristic-based densification with a probabilistic, adaptive mechanism.

Visual Acoustic Fields
Yuelei Li, Hyunjin Kim, Fangneng Zhan, Ri-Zhao Qiu, Mazeyu Ji, Xiaojun Shan, Xueyan Zou, Paul Liang, Hanspeter Pfister, Xiaolong Wang
CVPR 2025 3rd Workshop on Generative Models for Computer Vision
Paper / Website / Data / Code

We introduce Visual Acoustic Fields, a framework that bridges hitting sounds and visual signals within a 3D space using 3D Gaussian Splatting.


Qualcomm RB3 Color Cube Perception and Grasping
Hyunjin Kim*, Woojeh Chung*, Jeongeun Lee*
Supervised by Henrik I. Christensen, Sriraj Kandala and Ramakrishna Kintada
Qualcomm University Relations Platforms Symposium, 2025
Project Video / Code / Linkedin Post

We built an end-to-end manipulation pipeline on Qualcomm's RB3 platform that achieves real-time, vision-guided pick-and-place through lightweight on-device perception, planning, and control.

Codebase for Qualcomm RB5 MBot Control
Hyunjin Kim*, Sohyun Yoo*
Projects for UCSD CSE 276A: Introduction to Robotics, 2024
Project Videos / Code

We developed a complete autonomous mobile robot system integrating EKF-SLAM for localization and mapping, path planning algorithms for obstacle avoidance, and behavior-based control for area coverage navigation.


M.S. in Computer Science | University of California San Diego
Sep. 2024 - June 2026 (expected)

B.S. in Urban Planning and Engineering, Computer Science and Engineering | Yonsei University
Mar. 2020 - Aug. 2024

  • Exchange Student at University of Toronto, 2022-23

  • Google Developer Student Clubs (GDSC) Solution Challenge Top 100 Finalist, 2024
  • Inclusion@RSS Fellowship, Robotics: Science and Systems (RSS), 2023
  • Honors, Yonsei University, 2021 & 2024
  • High Honors, Yonsei University, 2022
  • Jilli (Merit-Based) Scholarship, Yonsei University, 2021-22

  • Reviewer
    • Journal Reviewer: IEEE TIP (2025)

Instructor, QI AI Entrepreneurship Program, Qualcomm Institute (CalIT2), July-Aug 2025
Instructor, QI AI Research-focused Program, Qualcomm Institute (CalIT2), Jan-Feb 2026



Website based on the template here and here.