# G3T Up! Gravity Aligned Coordinate Frames Simplify Pointmap Processing 

[![Project Page](https://img.shields.io/badge/Project-Page-4CAF50?logo=googlechrome&logoColor=green)](https://g3t-paper.github.io/)
[![Paper](https://img.shields.io/badge/Paper-arXiv-b31b1b?logo=arxiv&logoColor=red)](https://arxiv.org/)
[![Code](https://img.shields.io/badge/Code-GitHub-181717?logo=github&logoColor=black)](https://github.com/g3t-paper/g3t/)

[Bharath Raj Nagoor Kani](https://bharathrajn.com/), [Noah Snavely](https://www.cs.cornell.edu/~snavely/) <br/>
Cornell University

## Overview

<p align="center">
  <img src="teaser.png" alt="teaser_img" />
</p>

<p align="center"> We introduce <strong>G3T</strong>, a transformer that predicts upright, gravity-aligned pointmaps regardless of input image orientation, and <strong>G3T-Long</strong>, a pipeline that leverages this uprightness to enable robust long-sequence 3D reconstruction. </p>

## Setup

Please refer to our [GitHub repository](https://github.com/g3t-paper/g3t/).

## Citation

If you find our work useful, please consider citing our paper:

```
@article{kani2026g3t,
  author    = {Nagoor Kani, Bharath Raj and Snavely, Noah},
  title     = {G3T Up! Gravity Aligned Coordinate Frames Simplify Pointmap Processing},
  journal   = {arXiv preprint},
  year      = {2026},
}
```