Janghyeon Lee

Korea Advanced Institute of Science and Technology, LG (South Korea)

Papers

Total Citations

111

H-Index

About

Janghyeon Lee is a computer vision researcher whose work focuses on a critical challenge for autonomous systems: how machines perceive depth from visual data. His research spans monocular depth estimation and depth completion, addressing the fundamental problem of reconstructing three-dimensional scene geometry from two-dimensional images and sparse sensor data. Lee’s most influential contribution is the “Patch-Wise Attention Network for Monocular Depth Estimation” (2021, 66 citations), which introduced a novel attention mechanism that significantly improved the quality of depth maps from single images—a capability essential for robotics and autonomous driving. He further advanced the field with his work on cross-guidance architectures that fuse single images with sparse LiDAR data for depth completion (2020, 42 citations), tackling the persistent challenge of noise and sparsity in laser scan data. His more recent research on multi-scaled, densely connected locally convolutional layers (2022) continues to push the boundaries of depth completion accuracy. Lee’s contributions are particularly impactful for downstream tasks in autonomous vehicles and robot vision, where reliable depth perception is critical for safe navigation and scene understanding.