Gwangtak Bae

I'm a Ph.D. student at 3D Vision Lab, advised by Prof. Young Min Kim at Seoul National University (SNU). Before joining the lab, I had a great time  as a Research Officer in Agency for Defense Development (ADD). I received a B.S. in Electrical Engineering from Korea Institute of  Science and Technology (KAIST). I'm interested in designing robust vision systems for robots.

• Contact: tak3452@{snu.ac.kr, gmail.com}

• Curriculum Vitae: [CV] (updated: Jul. 2025)

Learning  3D Scene Analogies with Neural Contextual Scene Maps

Junho Kim, Gwangtak Bae, Eunsun Lee, and Young Min Kim

International Conference on Computer Vision (ICCV), 2025 

[arXiv]

We propose a new task of finding 3D scene analogies, which are dense maps connecting regions sharing similar scene contexts.

I²-SLAM: Inverting Imaging Process for Robust Photorealistic Dense SLAM

Gwangtak Bae*, Changwoon Choi*, Hyeongjun Heo, Sang Min Kim, and Young Min Kim

(*Equally contributed)

European Conference on Computer Vision (ECCV), 2024 

[arXiv] [paper] [supp]

In this paper, we invert imaging process to improve robustness and accuracy of dense SLAM in real-

world data which frequently contains motion blur and varying appearances.

SLiDE: Self-supervised LiDAR De-snowing through Reconstruction Difficulty

Gwangtak Bae, Byungjun Kim, Seongyong Ahn, Jihong Min, and Inwook Shim

European Conference on Computer Vision (ECCV), 2022 

[arXiv] [paper] [supp]

We propose a self-supervised LiDAR de-noising method that removes noise points in snowy weather,

which is one of the biggest challenges for 3D perception in autonomous driving.