My research develops physically grounded 3D representations and rendering
algorithms for inverse rendering, Monte Carlo reconstruction, and generative 3D content. I am currently interested in developing 3D world generation framework that combines visual
generation with physically grounded scene properties for editing and simulation. My broader goal is to move beyond static appearance modeling toward physically-grounded and interactive 3D worlds.
qbhan@kaist.ac.kr,
qbhan98@gmail.com,
smarthkb98@gmail.com, 
Research Interest
Physically Based Rendering, Neural Rendering, 3D Generation, Inverse Rendering
News
•
Jun 2026 - Our work(2nd author) has been accepted to ECCV 2026!
•
Feb 2026 - RadioGS has been accepted as an Oral paper for ICLR 2026!!
•
•
Nov 2025 - I will be participating at ICCV 2025! See you guys there!!
•
Aug 2025 - Our work has been accepted to BMVC 2025!!
•
May 2025 - Our work has been accepted to ICIP 2025!!
•
Nov 2024 - I will be participating at SIGGRAPH ASIA 2024! See you guys there!!
•
Oct 2024 - I’ve been acknowledged by Krafton’s recent neural rendering paper which aims to accelerate the optimization using semi gradients!
•
Mar 2024 - Our work has been accepeted to CVPR 2024!!
•
Aug 2023 - I’m pursuing Ph.D. in our SGVR Lab!
•
Jul 2023 - I got a Best Master's Thesis Award from Korea Computer Graphics Society! Thank you for the honor!
More…
Education
•
PhD, Computer Science, KAIST (Aug 2023 ~ Present)
◦
GPA 4.23/4.3
•
MS, Computer Sciene, KAIST (Aug 2021 ~ Aug 2023)
◦
GPA 4.2/4.3
•
BS, Computer Science, Minor in Electrical Engineering, KAIST (Mar 2017 ~ Aug 2021)
◦
Cum Laude (GPA 3.80/4.3)
•
Korea Science Academy of KAIST (Feb 2014 ~ Feb 2017)
Publications
•
What You Ask is What You Ground: Bridging Question Intent to Temporal Evidence for Grounded VideoQA
ECCV 2026
•
ICLR 2026 Oral (223 of 19,814 submissions; 1.13%)
•
AegisRF: Adversarial Perturbations Guided with Sensitivity for Protecting Intellectual Property of Neural Radiance Fields
BMVC 2025
•
Pose-free 3D Gaussian splatting via shape-ray estimation
ICIP 2025 (Best Student Paper Award)
Youngju Na*, Taeyeon Kim*, Jumin Lee, Kyu Beom Han, Woo Jae Kim, Sung-Eui Yoon
•
UFORecon: Generalizable Sparse-View Surface Reconstruction from Arbitrary and UnFavOrable SEts
CVPR 2024
•
•
User-Controlled Layout Editing with Neural Style Transfer
Proceedings of Korea Computer Grahpics Society Conference (KCGS) 2023
•
Monte Carlo Image Denoising using Spatial Information of Light Bounces
Proceedings of Korea Information Science Society Conference (KCC) 2022
Invited Talks & Writings
•
Radiometrically Consistent Gaussian Surfels for Inverse Rendering
KCCV 2026
•
Pixel-wise Guidance for Utilizing Auxiliary Features in Monte Carlo Denoising
KCGS 2023
•
Computer Graphics: Creating a New World Inside the Computer
KAIST magazine “Behind Science” 2023
•
Accelerating Deep-Learning-Based Denoising Methods via High Performance Computing
GPU tutorial of KSC 2022
Scholarships and Awards
•
NRF Ph.D. Fellowship from National Research Foundation of Korea (2024~2025)
•
Best Master's Thesis Award from Korea Computer Graphics Society (2023)
•
Excellence Award as Teaching Assistant in School of Computing, KAIST (2023 Spring, 2022 Fall, 2022 Spring)
•
Excellence Award in Undergraduate Research Program in KAIST (2021)
•
Korea Presidential Science Scholarship ($10K per year) (2017~2020)
•
Korea National Representative of APAO (Asia-Pacific Astronomy Olympiad) (2013, 2015)
Won a Silver Medal in APAO 2013 Junior session
Experience
Teaching
•
Teaching Assistant, Computer Graphics (CS500), Fall 2023
•
Teaching Assistant, Data Structure (CS206), Spring 2023, Fall 2022, Spring 2022
•
Teaching Assistant, Interactive Computer Graphics (CS482), Fall 2021
Projects
•
Kernel Refinement for Monte Carlo Denoising using Pixel-wise Discriminator (2022)
Class Project (CS580)
◦
Voted as Best Project
◦
Refine denoising kernel based on pixel-wise scores of the U-Net discriminator
•
Indoor Spatial Modeling and Rendering via Neural Rendering (2021)
Research Project
•
Contribution of Auxiliary Features to Monte Carlo Denoisers Based on Deep Learning (2021)
Research Project (Excellence Award on Undergraduate Research Program in 2021)
◦
Analyzing & Enhancing contributions of auxiliary features (albedo, normal, and depth) by adding channel attention to existing denoisers
◦
Applied multi-task learning by adding auxiliary tasks of reconstructing auxiliary features from a denoised image to enhance the semantics of auxiliary features
•
Scene Generator with Blender 2.93 (2021)
Research Project
◦
Generates scenes with various camera, material, and lighting settings by path-tracing using Cycles renderer supported by Blender
◦
Extracts geometric features (albedo, normal, depth) of each scene
◦
Current Monte-Carlo denoising methods involves deep learning which requires a lot of data. This scene generator is expected to make rich training dataset for this task easily.
•
Simple Ray Tracer on CPU (2020)
Individual Project
◦
Re-implemented the ray tracer following the book series Ray Tracing in One Weekend
•
Motion Detector for Interactive Online Real-time Class (2020)
Class Project (Introduction to AI CS470, A+)
◦
Detects simple hand gestures and alarms sudden movement for interactive online classes
◦
Contributing to preprocessing data and implementing an application using Openpose and Zoom SDK
•
Mini-C Interpreter (2019)
Class Project (Compiler Design CS420)
◦
Implemented a mini-C interpreter with an error handler that handles common syntax errors
◦
Contributed to constructing syntax rules of mini-C and implementing lexer and parser
•
MADMOVIE, a movie-critics website (2019)
Individual Project
◦
Implemented and managed server and database based on Node.js and PostgreSQL
◦
Crawled information and critics of currently screening movies using Beautiful Soup and Python
