# G3T Up! Gravity Aligned Coordinate Frames Simplify Pointmap Processing

Source: [G3T Up! Gravity Aligned Coordinate Frames Simplify Pointmap Processing](<https://www.sota2.com/research/paper/6a167adbdbd7150506f8033c>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-28T05:00:13.160Z

Authors: Bharath Raj Nagoor Kani, Noah Snavely

Published: 2026-05-26T17:59:55.000Z

## Abstract

Modern feed-forward 3D reconstruction methods like VGGT predict pixel-aligned pointmaps in camera-centric coordinate frames. However, this choice of coordinate frame is not always optimal. We propose instead to predict pointmaps in upright, gravity-aligned frames that exploit strong structural cues present in many real-world scenes. Unlike camera-centric frames, gravity-aligned frames share a common vertical axis across viewpoints, reducing the rotational degrees of freedom needed to relate pointmaps to one another. To this end, we introduce the Gravity Grounded Geometry Transformer (G3T), fine-tuned from existing models on gravity-aligned 3D data. G3T produces highly accurate gravity-aware predictions, including upright pointmaps and camera-to-gravity poses. We further introduce G3T-Long, a submap-based incremental 3D reconstruction pipeline that leverages the reduced rotational degrees of freedom afforded by upright frames to achieve significantly improved reconstruction accuracy.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Camera-to-gravity rotation estimation on Hypersim (held out scenes)](<https://www.sota2.com/research/sota/camera-to-gravity-rotation-estimation-on-hypersim-held-out-scenes.md>): Camera-to-gravity rotation estimation; Hypersim (held out scenes); Mean Rotation Error: 0.58.

- [Camera-to-gravity Rotation Estimation on TartanAir (held out scenes)](<https://www.sota2.com/research/sota/camera-to-gravity-rotation-estimation-on-tartanair-held-out-scenes.md>): Camera-to-gravity rotation estimation; TartanAir (held out scenes); Mean Rotation Error (Rerr): 1.41.

- [Camera-to-gravity rotation estimation on 7Scenes](<https://www.sota2.com/research/sota/camera-to-gravity-rotation-estimation-on-7scenes.md>): Camera-to-gravity rotation estimation; 7Scenes; Mean Rotational Error (Rerr Mean): 1.78.

- [Camera-to-gravity rotation estimation on NRGBD](<https://www.sota2.com/research/sota/camera-to-gravity-rotation-estimation-on-nrgbd.md>): Camera-to-gravity rotation estimation; NRGBD; Rotation Error (Mean): 1.13.

- [Camera-to-gravity Rotation Estimation on ETH3D](<https://www.sota2.com/research/sota/camera-to-gravity-rotation-estimation-on-eth3d.md>): Camera-to-gravity rotation estimation; ETH3D; Mean Rotational Error: 1.05.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2605.27372>)

[PDF](<https://arxiv.org/pdf/2605.27372.pdf>)
