metadata
license: apache-2.0
pipeline_tag: image-to-3d
R³: 3D Reconstruction via Relative Regression
R³ is a feed-forward geometry foundation model that reconstructs camera poses and dense geometry from arbitrarily long video streams via relative-pose regression. Instead of regressing every camera in one global frame, R³ predicts confidence-weighted pairwise relative poses on top of a Depth Anything 3 backbone, then assembles a consistent global trajectory downstream.
Project Page | Paper | GitHub
Quick Start
Installation
conda env create -f environment.yml
conda activate r3
pip install -e .
Alternatively, if you already have a CUDA-enabled PyTorch environment:
pip install -r requirements.txt
pip install -e .
Run the Demo
You can run inference on a sequence and visualize it using the following command:
python demo.py --seq_path examples/indoor --no_viewer
Presets for common scenarios:
python demo.py --mode local # indoor scenes, small coverage
python demo.py --mode long # long trajectories, large outdoor scenes
python demo.py --mode strided # temporally strided video
Citation
If R³ is useful in your research or projects, please cite:
@article{r3_2026,
title = {R^3: 3D Reconstruction via Relative Regression},
author = {Anonymous},
year = {2026},
note = {Paper coming soon}
}