Conference and workshop presentations, including IEEE INDIN 2026 and ECML PKDD 2025.
industrial ai / machine learning / automation
Research
I work at the intersection of machine learning research and real-world deployment, with projects spanning 3D human pose estimation, graph neural networks, computer vision, forecasting, anomaly detection, and intelligent industrial systems.
Expected PhD graduation: February 25, 2027
summary
Industry and government projects in energy, emissions, air quality, and inspection.
5 registered software works and 2 patent applications.
Research, mentoring, and software development at Chosun University.
conference and workshop presentations
conference presentation
Transformer-Based Occlusion for 3D Human Pose Estimation: A Focus on Reconstruction of Invisible Keypoints
A Transformer-based computer-vision framework for reconstructing occluded skeletal keypoints in 3D human pose estimation, improving robustness in complex visibility conditions. Presented at two venues.
Authors: Duncan Kibet, Minseop So, HaHyeon KangAdvisor: Prof. Jong-Ho Shin
poster presentation
Development of an AI-Based Elbow Range-of-Motion Measurement and Classification Model
A webcam-based AI pipeline that extracts 3D shoulder, elbow, and wrist landmarks to measure dynamic elbow range of motion and classify movement status, combining MediaPipe Pose, quality control, feature engineering, and a Random Forest classifier for real-time assessment.
Authors: Ho-Yong Park, Seung-Ok Lee, Geun-Young Park, Ha-Yeon Kang, Duncan Kibet, Jong-Ho Shin
poster presentation
Advancing ATC Drug Classification Through Graph Convolutional Networks
Improving Anatomical Therapeutic Chemical (ATC) drug classification through graph convolutional networks and molecular graph representation learning.
workshop presentation
Graph Convolutional Networks for ATC Drug Classification from Molecular Data
Molecular representation learning for ATC prediction, presented at the 5th PharML Workshop at ECML PKDD 2025.
Authors: Duncan Kibet, Min-Seop So, Hahyeon Kang, Prof. Jong-Ho Shin
paper presentation
Enhanced Time Series-Based Solar Power Prediction Using Minimal Feature Selection
Solar power forecasting using feature selection and Transformer-based modeling, presented at the TempXAI Workshop at ECML PKDD 2025.
Authors: Duncan Kibet, Min-Seop So, Hahyeon Kang, Prof. Jong-Ho Shin
conference research
Research on Fall Detection Using Transformer Model Based on Video Data
A video-based fall detection framework using Transformer models for safety monitoring.
Authors: Ha Hyun Kang, Min Seop So, Duncan Kibet, Prof. Jong-Ho Shin
conference research
Machine Learning-Based Anomaly Detection in Optical Lenses
Anomaly detection techniques for optical lens inspection.
Authors: Duncan Kibet, Prof. Jong-Ho Shin
conference research
Development of AI-Based Fall Detection Model Using Transformers
Model architecture design and preprocessing for a Transformer-based fall detection system.
funded and industry projects
funded project
AI Model for Predicting Power Plant Emissions
Predictive models for SOx and NOx emissions from coal-fired power plants, supported by the National Research Foundation of Korea.
funded project
AI Value Assessment for Solar Power Plants and Advanced Fault Prediction
AI models for solar asset valuation and fault diagnosis.
funded project
AI Model for Solar Power Forecasting Using Time-Series Data
Transformer-based time-series forecasting for solar power generation.
funded project
AI Model for Air Quality Prediction and Control
Fine dust and indoor air quality prediction with deep learning (LSTM, CNN).
applied vision
Image Classification on Lens Images
A ResNet-50 and Grad-CAM based inspection system for normal and abnormal lens images.
laboratory research
Detecting Anomalies in Solar Power Plants
XGBoost, AE-LSTM, and Isolation Forest models for anomaly detection in operational energy systems.