siyan824/profusion_pr

PROFusion: Robust and Accurate Dense Reconstruction via Camera Pose Regression and Optimization

[ICRA 2026] PROFusion is a simple yet effective system for real-time camera tracking and dense scene reconstruction, providing both robustness against unstable camera motions and accurate reconstruction results.

This repository contains pre-trained weights for the pose regression module, which estimates the relative camera pose (in metric-scale) between two RGB-D frames.

• Paper: PROFusion: Robust and Accurate Dense Reconstruction via Camera Pose Regression and Optimization
• Code: GitHub Repository

Method Overview

Real-time dense scene reconstruction during unstable camera motions is crucial for robotics, yet current RGB-D SLAM systems often fail when cameras experience large viewpoint changes, fast motions, or sudden shaking. PROFusion addresses this challenge through a combination of learning-based initialization with optimization-based refinement.

The system employs a camera pose regression network to predict metric-aware relative poses from consecutive RGB-D frames, which serve as reliable starting points for a randomized optimization algorithm that further aligns depth images with the scene geometry. The system operates in real-time, showcasing that combining simple and principled techniques can achieve both robustness for unstable motions and accuracy for dense reconstruction.

Citation

If you find this work helpful in your research, please consider citing:

@article{dong2025profusion,
  title={PROFusion: Robust and Accurate Dense Reconstruction via Camera Pose Regression and Optimization},
  author={Dong, Siyan and Wang, Zijun and Cai, Lulu and Ma, Yi and Yang, Yanchao},
  journal={arXiv preprint arXiv:2509.24236},
  year={2025}
}

Acknowledgments

The implementation is based on several inspiring works in the community, including DUSt3R, SLAM3R, Reloc3r, and ROSEFusion.