Junhong Min’s homepage
I am a Staff Research Scientist at Google, Intrinsic, based in Mountain View, California. My research focuses on physical intelligence and 3D vision, with an emphasis on stereo matching, depth estimation, and robotic manipulation.
Previously, I was a Senior Research Scientist at Holiday Robotics and a Principal Engineer at Samsung Electronics. My work spans robust stereo depth estimation in transparent, reflective, and cluttered scenes, 6-DoF object pose estimation, and vision systems for industrial automation. Earlier research explored computational imaging for super-resolution microscopy and low-dose CT reconstruction.
Featured Publications
- Comprehensive Survey on Advances and Challenges in RGB-D Semantic Segmentation, Pattern Recognition 2026. Paper / GitHub
- DepthFocus: Controllable Depth Estimation for See-Through Scenes, CVPR 2026. Paper / Project
- S²M²: Scalable Stereo Matching Model for Reliable Depth Estimation, ICCV 2025. Paper / Project / Code
- Confidence Aware Stereo Matching for Realistic Cluttered Scenario, ICIP 2024. Paper
- Sim-to-real Object Pose Estimation for Random Bin Picking, ICRA 2024. Paper
View full publication list / View patents
Latest News
Nov 2026: Joined Google, Intrinsic as a Staff Research Scientist in Mountain View, California.
Jul 2026: Joined Holiday Robotics as a Senior Research Scientist.
May 2026: Our survey, “Comprehensive Survey on Advances and Challenges in RGB-D Semantic Segmentation,” was accepted to Pattern Recognition.
Apr 2026: Won the CVPR NTIRE 2026 stereo track for HR depth from specular and transparent surfaces.
Feb 2026: Our paper, “DepthFocus: Controllable Depth Estimation for See-Through Scenes” was accepted to The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2026!
Oct 2025: Our paper, “S²M²: Scalable Stereo Matching Model for Reliable Depth Estimation” was accepted to the International Conference on Computer Vision (ICCV) 2025!
