Eojindl Yi

Korea Advanced Institute of Science and Technology

Papers

Total Citations

H-Index

About

Eojindl Yi is a researcher advancing the frontier of 3D scene understanding, with a primary focus on monocular depth estimation and depth completion. Yi’s work addresses a critical challenge in computer vision: extracting reliable depth information from a single RGB image, a capability essential for autonomous vehicles, robotics, and mobile systems. Their most influential contribution, the “Patch-Wise Attention Network for Monocular Depth Estimation” (2021), has garnered 66 citations, demonstrating significant impact in the field. This work introduced a novel attention mechanism that enables a network to focus on local image patches, improving the quality of predicted depth maps. Building on this, Yi developed the “Token-Sharing Transformer” (2023), a lightweight architecture designed for resource-constrained mobile robotics, achieving efficient depth estimation without sacrificing accuracy. In the domain of depth completion—fusing sparse LiDAR data with RGB images to produce dense depth maps—Yi proposed “Multi-Scaled and Densely Connected Locally Convolutional Layers” (2022). This architecture leverages multi-scale features and dense connections to enhance the fidelity of completed depth maps, directly supporting downstream tasks in autonomous navigation. Through these contributions, Yi is helping to make robust 3D perception more accessible and deployable in real-world systems.