Donghwan Lee
Papers
2
Total Citations
33
H-Index
2
About
Donghwan Lee is a robotics researcher whose work advances autonomous navigation in complex, real-world environments. His primary research areas include 3D deep learning, place recognition, and self-supervised depth estimation for robotic perception. Lee’s most impactful contribution is **SpoxelNet**, a novel framework for place recognition in crowded indoor spaces that processes 3D point clouds using spherical voxel representations. This work, which has garnered **29 citations**, addresses a critical gap in robotics by enabling robust localization in cluttered, feature-sparse environments where traditional methods fail. More recently, Lee introduced **SelfTune**, a self-supervised learning algorithm that resolves the scale ambiguity inherent in monocular depth estimation. By integrating monocular SLAM with proprioceptive sensor data, SelfTune produces metrically accurate depth maps without requiring ground-truth labels—a significant step toward practical, low-cost autonomous systems. Lee’s research is notable for its focus on making deep learning-based perception reliable in unstructured, real-world settings, directly supporting the goal of full robot autonomy.
Research Focus
Key Achievements
Top Papers
- 1
- 2