Kangmin Seo

Kangmin Seo

Suwon, South Korea • kangrnin@gmail.com • LinkedIn • GitHub
M.S. Student at Visual Computing Lab @ SKKU,
Focusing on Efficient and Robust 3D Representations.

Publications

FlatMinGS figure
Preprint
Improving Sparse-View 3DGS Generalization via Flat Minima Optimization
Kangmin Seo, Sangeek Hyun, MinKyu Lee, Jae‑Pil Heo
Preprint (Under Review)
PDF-GS figure
CVPR Findings 2026
PDF-GS: Progressive Distractor Filtering for Robust 3D Gaussian Splatting
Kangmin Seo, MinKyu Lee, Tae‑Young Kim, ByeongCheol Lee, JoonSeoung An, Jae‑Pil Heo
IEEE/CVF Conference on Computer Vision and Pattern Recognition (Findings), 2026
arXiv Project Page Code
DFFSplat figure
NeurIPS 2025
Diffusion Feature Field for Text-based 3D Editing with Gaussian Splatting
Eunseo Koh, Sangeek Hyun, MinKyu Lee, Jiwoo Chung, Kangmin Seo, Jae‑Pil Heo
Neural Information Processing Systems, 2025
NeurIPS OpenReview

Education

Sungkyunkwan University (SKKU) Sep. 2025 – Present
M.S. Student in Software
At Visual Computing Lab (VCLab), Advisor: Prof. Jae-Pil Heo
Sungkyunkwan University (SKKU) Mar. 2019 – Aug. 2025
B.S. in Software
GPA: 4.22 / 4.5 (Major: 4.37 / 4.5), Magna Cum Laude (Top 2.9%)
Korea Digital Media High School Mar. 2016 – Feb. 2019
Hacking Defense Department

Experience

Visual Computing Lab (VCLab) @ SKKU Dec. 2024 – Present
M.S. Student (Advisor: Prof. Jae-Pil Heo) | Suwon, Korea
  • Specializing in 3D Computer Vision, with a focus on improving the efficiency and robustness of 3D Gaussian Splatting (3DGS).
Machine Learning and Intelligence Lab @ KAIST Jan. 2024 – Sep. 2024
Research Intern | Seoul, Korea (Remote)
  • Conducted experiments for an LLM-based patient interview service project.
  • Explored research ideas in LLMs, focusing on efficiency (offloading) and reasoning (test-time scaling)
RideFlux Inc. June 2023 – Aug. 2023
Perception SW Engineer Intern | Seoul, Korea
  • Developed a video-based object tracking PoC for autonomous vehicles (Center Track).
  • Improved robustness against motion blur during vehicle rotation via adaptive thresholding.
Republic of Korea Air Force May 2021 – Dec. 2022
3D/VR System Engineer (Military Service) | Daejeon, Korea
  • Developed a high-fidelity VR flight training simulator for pilot education using Unreal Engine.
  • Implemented complex avionics logic, motion tracking interfaces, and multiplayer synchronization.

Selected Projects

RealBang: 3D Room Generation AI & Viewer Mar. 2023 – Jan. 2024
Python, PyTorch, 3DGS, NeRF
  • Developed a pipeline to reconstruct 3D room geometry from smartphone videos using NeRF and 3DGS.
  • Addressed distortion issues of 2D real estate photos by providing immersive 360-degree 3D views.
  • Implemented a client-side viewer for seamless exploration of reconstructed 3D scenes.
  • Awarded Grand Prize (President's Award) at SKKU Co-Deep Learning Project.
AR Experience App for Busan One Asia Festival Apr. 2020 – Feb. 2021
Android, ARCore, YOLOv4
  • Developed an interactive AR content app using Android ARCore and YOLOv4 object detection.
  • Researched indoor localization using Wi-Fi fingerprinting and ML models.

Honors & Awards

Scholarships
  • Graduate Merit Scholarship (Full Tuition) Fall 2025 – Present
  • Student Success Scholarship (Merit Grant) Spring 2025
  • Academic Excellence Scholarship (Full Tuition) Fall 2023
  • Sungkyun Software Scholarship (Half Tuition) Spring 2019 – Fall 2020
Awards
  • 2024 SKKU Programming Contest – Bronze Prize Nov. 2024
  • SKKU Co-Deep Learning Project – Grand Prize (President's Award) Jan. 2024
  • 3rd University Alliance Hackathon (Uni-DTHON) – Silver Prize Nov. 2023
  • 2023 SKKU Programming Contest – Bronze Prize Mar. 2023
  • 2022 Military AI Contest (MAICON) – Grand Prize (Minister of Science and ICT Award)
    aitimes.kr     Nov. 2022