Jemo Maeng

I earned my M.S. at GIST and am now a Ph.D. student in the Department of AI there, advised by Prof. Kyoobin Lee at GIST AILAB. Before that I took my B.S. in Computer Science and Technology at Harbin Institute of Technology in China.

My work is about perception that keeps working once it leaves the lab.

Research interests

Multi-modal perception. A single camera fails predictably — fog, rain, glare, darkness, smoke. I fuse RGB, infrared and LiDAR so that detection and segmentation degrade gracefully instead of collapsing. Most of this is grounded in real deployments: drone views in GPS-denied environments, maritime scenes from unmanned surface vehicles, and mobile robots whose input distribution drifts with the weather.

Dexterous manipulation. Vision alone does not tell you whether a grasp is about to slip. I work on learning tactile representations that let a multi-fingered hand handle unstructured, previously unseen objects — no CAD model, no fixture.

Robustness under domain shift. Threading both: test-time and continual adaptation, so a model trained yesterday still holds up on today’s inputs.

Selected highlights

• 🥇 First Prize, Multimodal Semantic Segmentation Challenge — 4th Maritime Computer Vision Workshop, CVPR 2026
• 🏅 Minister’s Award (MOTIE), 3rd Electrical Safety Excellent Paper Competition, 2022
• 📄 ICRA 2025 — high-quality unknown object instance segmentation via quadruple boundary error refinement
• 🤖 Project lead on three funded programmes, including an amphibious reconnaissance robot (ICMTC) and a two-step test-time domain adaptation method for LG Electronics

The full record is on my CV; projects are broken out under Projects and papers under Publications.

Contact

Dasan Building 204, GIST, Gwangju, South Korea  ·  maengjemo@gm.gist.ac.kr