VG3T: Visual Geometry Grounded Gaussian Transformer

Paper | Project Page | Code

VG3T: Visual Geometry Grounded Gaussian Transformer

Junho Kim , Seongwon Lee$\dagger$

$\dagger$ Corresponding Author.

News

• [2026/1/31] VG3T is accepted to ICRA2026!
• [2025/12/28] Paper released on arXiv.

Demo

Overview

Getting Started

Code will be released upon publication.

Related Projects

Our code is based on the excellent work GaussianFormer.

Citation

If you find this project helpful, please consider citing the following paper:

@article{kim2025vg3tvisualgeometrygrounded,
  author    = {Junho Kim and Seongwon Lee},
  title     = {VG3T: Visual Geometry Grounded Gaussian Transformer},
  journal   = {IEEE International Conference on Robotics and Automation (ICRA)},
  year      = {2026},
}