Hyeonseo Nam

Hello, I am an AI researcher at the AI research division of a telecom company in Korea, focusing on leveraging Large Language Models (LLMs) to enhance workflow efficiency. My primary interest is in productivity enhancements. Professionally, I research the applications of LLMs, including semantic document parsing, Retrieval-Augmented Generation (RAG), and chatbots. Personally, I contribute to open-source projects aimed at improving productivity.

During my graduate studies, I researched the fields of digital human and 3D reconstruction, handling various data representations. Through this, I have participated in research papers at top-tier conferences in the graphics field and contributed to open-source projects.


namhs2030 [at] gmail.com / [github] / [google scholar] 

Education

Featured Activities

Plugin

Obsidian Plugin - Auto Classifier
A plugin for obsidian that automatically classify user's note using ChatGPT API
[repo] (2023)

Face Reconstruction

📌 Animatable 3D Head Reconstruction with Geometric Details from a Single Image
Hyeonseo Nam, Seungyoung Lee
Thesis, 2022
[pdf] 

We've trained Single-image-based deep implicit function for animatable face reconstruction with 3D face dataset. Our reconstructed model can express diverse facial expressions (using 3DMM property) because of our semantic-aware training scheme.

Human Reconstruction

📌 LaplacianFusion: Detailed 3D Clothed-Human Body Reconstruction
Hyomin Kim, Hyeonseo Nam, Jungeon Kim, Jaesik Park, Seungyong Lee
SIGGRAPH Asia 2022
[pdf] / [doi] / [video] [video2]

We’ve introduced a dynamic human reconstruction framework from depth or point cloud sequence. We used Laplacian coordinate to capture more surface details on SMPL quried surface function. 

Open Source

📌 Contributions to Trimesh open source
Python 3D data processing library, 2.2k stars
[PR: subdivision algorithm] / [PR: fix error] (2021-2022)

I implemented an advanced subdivision algorithm (loop subdivision method, def subdiviion_loop) and editted an error for mesh slicing function (def slice_mesh_plane).

All Activities

Publications - International

📌 Animatable 3D Head Reconstruction with Geometric Details from a Single Image
Hyeonseo Nam, Seungyoung Lee
Thesis, 2022
[pdf] 

📌 LaplacianFusion: Detailed 3D Clothed-Human Body Reconstruction
Hyomin Kim, Hyeonseo Nam, Jungeon Kim, Jaesik Park, Seungyong Lee
SIGGRAPH Asia 2022 (also published on ToG)
[pdf] / [doi] / [video] [video2]

TextureMe: High-quality Textured Scene Reconstruction in Real Time
Jungeon Kim, Hyomin Kim, Hyeonseo Nam, Jaesik Park, Seungyong Lee
ACM Transactions on Graphics 2022 (ToG)
[pdf] / [doi] / [video] [video2]

Spatiotemporal Texture Reconstruction for Dynamic Objects Using a Single RGB-D Camera
Hyomin Kim, Jungeon Kim, Hyeonseo Nam, Jaesik Park, Seungyong Lee
EuroGraphics 2021
[pdf]

Publications - Korean Domestic

정규화된 딥음함수를 이용한 단일 얼굴 영상 입력에서의 3D 얼굴 복원
Hyeonseo Nam, Seungyong Lee
IPIU 2022 (Best paper award - Silver prize) 

단일 영상 기반의 3D 인체 복원 연구 동향
Hyeonseo Nam, Seungyong Lee
KCC 2021

Patent

System and method for providing route guidance service for visually impaired people
US9970779, 2014
[US] / [KR]

Open Source

📌 Contributions to Trimesh open source
Python 3D data processing library, 2.2k stars
[PR: subdivision algorithm] / [PR: fix error] (2021-2022)

Obsidian Plugin - Emoji Titler
A plugin for obsidian that add an emoji to the title using keyboard shortcut
[repo] (2023)

Obsidian Plugin - Auto Classifier
A plugin for obsidian that automatically classify user's note using ChatGPT API
[repo] (2023)

ETC


Awards & Honor


ETC