Research Area
LoD-2 Building Model Reconstruction from Orthophotos and Point Clouds via 2D and 3D Break-Line Fusion: I investigated how 2D orthophoto information and 3D point-cloud data can be combined to reconstruct accurate building models. I focused on extracting and fusing important break-lines, including building boundaries, roof edges, and ridges, to improve roof geometry and overall reconstruction quality.
Federated Learning over Wireless Networks: I studied distributed machine-learning approaches that allow multiple wireless devices to collaboratively train a shared model without directly exchanging their raw data. I explored techniques such as model aggregation, communication-efficient training, and device coordination to understand how federated learning can be effectively deployed in wireless, IoT, edge-computing, and future 6G environments.
Quantum Machine Learning: I developed and experimented with parameterized quantum circuits and hybrid quantum-classical learning models using IBM Qiskit. I explored quantum data encoding, circuit design, parameter optimization, and quantum NN-based classification.
Disaster-Site Distance Estimation Using Monocular Depth Maps: I generated monocular depth maps from smartphone images using machine-learning methods, extracted depth information around disaster targets to estimate their distance from different camera positions, and evaluated the estimation accuracy by analyzing errors across different observation distances.