Joonkyu Min

I am an undergraduate student at Seoul National University, majoring in Electrical and Computer Engineering. I am currently an intern at CLVR Lab, where I work on robot safety under supervision of Joseph J. Lim.

Previously, I worked as an intern at SNU VGI Lab under professor Jaesik Park, where I worked on 3D computer vision, especially 3D gaussian splatting.

Email  /  GitHub  /  LinkedIn  /  CV

Research Goal

My goal is to build safe and reliable AI agent that can interact with the real world.

I believe that safety is the key component for practical tasks of robots. Safety gaurantees would also enable easier and scalable real world data collection and training, enabling more robust deployment!

Publications

CF3: Compact and Fast 3D Feature Fields

Proposed an approach for constructing a compact 3D feature representation from given 3D Gaussian Splatting

Projects

These include coursework, side projects and unpublished research work.

KL-regularized FB representation for offline zero-shot RL

Graduation project at SNU. Applied KL-divergence regularization approach to zero shot RL method (FB representation).

Gaussian Splatting in the dark

Course project of 3D Computer Vision at SNU. We presented a method to learn robust gaussian splatting to render realistic novel views from dark and blurry scenes. The proposed method initialize gaussians based on dense point tracking method instead of keypoint matching, and utilize the generated depth map for depth regularization.

Design and source code from Jon Barron's website