Junha Lee

Research Scientist at SqueezeBits.

I received my Ph.D. (2025) and M.S. (2021) in Computer Science and Engineering from POSTECH, where I was supervised by Prof. Minsu Cho at the Computer Vision Lab. I also received my B.S. (2019) in CSE from POSTECH.

My research interests lie in 3D visual perception for embodied intelligence. My work has progressed from 3D geometric understanding, spanning correspondence and registration (HighDimConv, DHVR) and radiance-field representations (PeRFception), to 3D vision-language models for open-vocabulary perception (Mosaic3D, Affogato, SpaCeFormer), to embodied manipulation, with language-driven grasping and embodied reasoning (DextER). I am now working on generative models for 3D motion forecasting and contact-aware dexterous grasping. My goal is to build agents that perceive, predict, and act in 3D, efficiently enough for real robots.

CV Email GitHub Scholar LinkedIn

News

Jul 2026 Recognized as an Outstanding Reviewer for ECCV 2026!
Jun 2026 Our paper Affogato is accepted to ECCV 2026!
May 2026 Our paper SpaCeFormer is accepted to ICML 2026!
Feb 2026 Joined SqueezeBits as a Research Scientist.
Feb 2026 Our paper DextER is accepted to CVPR 2026!

Experience

2026.02 — Present

Research Scientist | SqueezeBits

Working on the intersection of efficient AI & physical AI

2024.09 — 2025.09

Research Intern | NVIDIA (Remote)

Hosted by Chris Choy and Jan Kautz. Topic: Open-vocabulary 3D Foundation Models

2024.06 — 2024.08

Research Intern | NVIDIA (On-site, Santa Clara)

Hosted by Chris Choy and Jan Kautz. Topic: 3D Multimodal Large Language Models

Education

2021 — 2025

Ph.D. in Computer Science and Engineering | POSTECH

Thesis: Learning Language-Driven 3D Perception and Manipulation via Vision-Language Models

2019 — 2021

M.S. in Computer Science and Engineering | POSTECH

Thesis: Global Point Cloud Registration using High-dimensional ConvNets and Hough Voting

2012 — 2019

B.S. in Computer Science and Engineering | POSTECH

Selected Publications

* Equal contribution

ECCV 2026

Affogato: Open-Vocabulary Affordance Grounding with Automatic Data Generation at Scale

*Junha Lee, *Eunha Park, Chunghyun Park, Dahyun Kang, and Minsu Cho Paper ·Project

ICML 2026

SpaCeFormer: Fast Proposal-Free Open-Vocabulary 3D Instance Segmentation

Chris Choy, Junha Lee, Chunghyun Park, Minsu Cho, Jan Kautz Paper ·Project ·Code

CVPR 2026

DextER: Language-driven Dexterous Grasp Generation with Embodied Reasoning

Junha Lee, Eunha Park, and Minsu Cho Paper ·Project ·Code

CVPR 2025

Mosaic3D: Foundation Dataset and Model for Open-Vocabulary 3D Segmentation

*Junha Lee, *Chunghyun Park, Jaesung Choe, Yu-Chiang Frank Wang, Jan Kautz, Minsu Cho, and Chris Choy Paper ·Project ·Code

ICCV 2021

Deep Hough Voting for Robust Global Registration

Junha Lee, Seungwook Kim, Minsu Cho, and Jaesik Park Paper ·Project ·Code

CVPR 2020 Oral

High Dimensional Convolutional Networks for Geometric Pattern Recognition

Christopher Choy, Junha Lee, Rene Ranftl, Jaesik Park, and Vladlen Koltun Paper ·Project ·Code

View all publications →

Services

Conference Reviewer: CVPR (2022-2026), ECCV (2022, 2024, 2026), ICCV (2021, 2023, 2025), ICLR (2026), BMVC (2021)
Journal Reviewer: TPAMI (2022, 2024, 2025), IJCV (2024), IEEE RA-L/ICRA (2021)