A Thought

Researcher · NAVER LABS

Siyeong Lee.

Learning to see.
Understanding the world.

My work spans computer vision, 3D perception, and representation learning, with interests in agent-centric perception and multimodal foundation models.

Google Scholar ↗GitHub ↗

01 / Research

From pixels
to perception.

01

3D & agent-centric perception

LiDAR understanding, long-tailed object detection, and robust learning under changing domains and label spaces.

02

Multimodal & representation learning

Multimodal foundation models, self-supervised learning, and the information-theoretic foundations of neural estimation.

03

Computational imaging

High dynamic range imaging, inverse tone mapping, and deep image restoration.

02 / Publications

Selected work

Full publication list ↗

8 selected publications

03 / Open source

Research into code.

More on GitHub ↗

INFORMATION THEORY / UAI 2022

Deconstructing MINE ↗

Official implementation of regularized mutual information-based losses for more stable neural estimation.

COMPUTATIONAL IMAGING / ECCV 2018

Deep Recursive HDRI ↗

Code for inverse tone mapping with generative adversarial networks, reconstructing HDR images from LDR inputs.