Jaeah Lee

I'm an incoming DPhil student in Computer Science at the University of Oxford, where I will be supervised by Christian Rupprecht and Ronald Clark. I also work as a Research Engineer Intern at KRAFTON AI and plan to graduate from Seoul National University with an M.S. in Artificial Intelligence in August 2026. Prior to this, I received dual B.S. degrees in Chemistry Education and Computer Science and Engineering from the same university.

My primary research interests span computer graphics and computer vision, including but not limited to neural representations and generative models. In particular, I explore geometry-aware generative modeling, as well as physically grounded methods for interactive 3D reconstruction and manipulation. Ultimately, my goal as a researcher is to help bring our inner thoughts and imagination to life without limitations.

I welcome opportunities to collaborate on new ideas. Please feel free to reach out if you're interested in my research!

Research

(*denotes equal contribution.)

Vitality-Aware Compression for Efficient Image-to-Shape Diffusion Transformers Jaeah Lee*, Hyunjin Kim*, Jaewoong Cho, and Gihyun Kwon ECCV, 2026 We propose the first approach to compress Diffusion Transformers (DiTs) for lightweight image-to-shape generation. project page (coming soon)  /  paper

Articulated Object Reconstruction from Rest-State Observation Daeun Lee, Jaeah Lee*, Woosung Kim*, Haebeom Jung, and Jaesik Park ECCV, 2026 We present Rest2Art, an approach for reconstructing articulated objects with part-level geometry and joint parameters using only rest-state observations. project page  /  paper (coming soon)

Recovering Dynamic 3D Sketches from Videos Jaeah Lee, Changwoon Choi, Young Min Kim, and Jaesik Park CVPR, 2025 We introduce Liv3Stroke, a novel approach for reconstructing 3D sketches of dynamic objects from video frames by using the deformability of each stroke. project page  /  paper  /  code

Improving Editability in Image Generation with Layer-wise Memory Daneul Kim, Jaeah Lee, and Jaesik Park CVPR, 2025 We propose a framework to improve editability in sequential image generation by focusing on the maintenance of existing contents at each step. project page  /  paper  /  code

3Doodle: Compact Abstraction of Objects with 3D Strokes Changwoon Choi, Jaeah Lee, Jaesik Park, and Young Min Kim SIGGRAPH (ACM Transactions on Graphics), 2024 We propose a novel approach for compact representation of objects from multiple views, utilizing a sparse set of strokes within 3D space. project page  /  paper  /  code

Experiences

KRAFTON AI Research Engineer Intern Main research area: 3D generative models

Jul. 2025 - Aug. 2026

3D Vision Lab, Seoul National University Research Intern Main research area: Non-photorealistic rendering

Jul. 2023 - Feb. 2024

Vision & Learning Lab, Seoul National University Research Intern Main research area: Neural rendering, Inverse graphics networks

Jun. 2022 - May. 2023

Biointelligence Lab, Seoul National University Internship in Undergraduate Research Opportunities Program (UROP) Main research area: Human-computer interaction, Object detection

Jul. 2021 - Sep. 2021