Research — Jong Gwang Kim

Research

Redevelopment design sits at the intersection of hundreds of households and the fabric of the city — yet much of its early-stage review still relies on repetitive manual work. Having experienced this gap firsthand while automating site plan generation in practice, I turned to Graph Neural Networks for an answer. When buildings and cities are read as graphs of nodes and relationships, machines can support a designer's judgment — from classifying BIM data to providing urban-context feedback at the earliest stages of design. My goal is to bridge hands-on redevelopment practice with AI research: building tools that let designers focus on better decisions.

Publications

Performance Comparison of Supervised and Self-Supervised Learning Graph Neural Networks for BIM Element Classification — teaser figure

2026Master's Thesis

BIM 요소 분류를 위한 지도 및 자기지도 그래프 신경망 성능 비교

Performance Comparison of Supervised and Self-Supervised Learning Graph Neural Networks for BIM Element Classification

Jong Gwang Kim, Hyeoncheol Kim

GNNBIMSelf-Supervised Learning

Figures

BIM model and its graph representation — BIM to Graph conversion (Autodesk Sample)

Line chart comparing HPO performance across models — HPO performance across models and configurations

Model performance curves under varying label budgets — Model performance by label budget

2025Conference

GNN 기반 건축설계 초기 단계 도시 맥락 피드백 프레임워크

A GNN-based Urban Context Feedback Framework for Early-Stage Architectural Design

Jong Gwang Kim, Hyeoncheol Kim

Summer Conference of the Society for Computational Design and Engineering, 2025

GNNUrban ContextEarly-Stage Design

Reports

기술 리포트 및 프로젝트 문서는 준비되는 대로 이곳에 정리할 예정입니다.