Ph.D. candidate Sungjae Shin, Master candidates Wanhee Kim, and Jinsung Choi from Prof. Hyun Myung’s laboratory in the School of Electrical Engineering at KAIST has achieved an outstanding accomplishment by receiving the Best Paper Award at the 25th International Conference on Control, Automation, and Systems (ICCAS 2025), organized by the Institute of Control, Robotics, and Systems (ICROS).

This year’s conference, held in Incheon Sondo ConvensiA during Nov. 4 – 7, featured 800 participants and 487 presented papers, among which five were selected for the Outstanding Paper Award Finalist. Among these, only one paper was ultimately honored with the Best Paper Award.
〈Illustration of the proposed system “GSDB”〉
The award-winning paper, titled “GSDB: A Lightweight Database for Gaussian Splatting Map-based Visual Localization Leveraging Edge-aware and Quality-guided View Filtering,” introduces a novel lightweight database construction framework that achieves over 90% size reduction by applying edge-aware and quality-guided view filtering to the learned Gaussian Splatting map. Experimental results demonstrate that the proposed approach maintains localization accuracy while significantly improving estimation speed compared with conventional approaches.

---

명현 교수 연구실 신성재 박사과정, 김완희, 최진성 석사과정이 2025년 제어로봇시스템학회(ICROS)에서 주최하는 국제학술대회 ICCAS 2025에서 최우수논문상을 수상하는 성과를 거두었습니다.

인천 송도 컨벤시아에서 11/4~7에 개최된 이번 국제학술대회에는 31개국에서 800여명이 참가하고, 총 487편의 논문이 발표되었으며, 이 중 우수논문상 Finalist로 5편이 선정되었고, 그 중 1편이 최우수논문상(Best Paper Award)으로 최종 선정되었습니다.

수상 논문은 “GSDB: A Lightweight Database for Gaussian Splatting Map-based Visual Localization Leveraging Edge-aware and Quality-guided View Filtering”로, 학습된 Gaussian Splatting 지도로부터 엣지 인식·품질 유도 뷰 선별을 통해 약 90% 이상 경량화된 데이터베이스를 구축하고, 정확도는 유지하면서 추정 속도는 기존 대비 대폭 향상됨을 실험으로 입증한 연구입니다.

Source: KAIST EE, Robot News