Mast3r Slam with Rerun

An unofficial implementation of MASt3R-SLAM: Real-Time Dense SLAM with 3D Reconstruction Priors

Uses Rerun to visualize, Gradio for an interactive UI, and Pixi for a easy installation

Installation

Using Pixi

Make sure you have the Pixi package manager installed

This is Linux only with an NVIDIA GPU.

Installation can take quite a while (10+ minutes) as it requires building, lietorch + curope + asmk + mast3r-slam matching kernels

git clone https://github.com/rerun-io/mast3r-slam.git
cd mast3r-slam
pixi run example-base

All commands can be listed using pixi task list

Usage

Gradio App

pixi run app

CLI

with pixi example task, this is the default that is more accurate

pixi run example-base

this is the fast version that is less accurate, but much faster

pixi run example-fast

You can see all tasks by running pixi task list

with python in pixi shell

python tools/mast3r_slam_inference.py --dataset data/normal-apt-tour.MOV --img-size 512 --config config/base.yaml

Acknowledgements

Thanks to the original Mast3r Slam, Mast3r, and Dust3r repos!

Mast3r-SLAM

@article{murai2024_mast3rslam,
    title={{MASt3R-SLAM}: Real-Time Dense {SLAM} with {3D} Reconstruction Priors},
    author={Murai, Riku and Dexheimer, Eric and Davison, Andrew J.},
    journal={arXiv preprint},
    year={2024},
}      

Mast3r

@misc{mast3r_arxiv24,
      title={Grounding Image Matching in 3D with MASt3R}, 
      author={Vincent Leroy and Yohann Cabon and Jerome Revaud},
      year={2024},
      eprint={2406.09756},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

Dust3r

@inproceedings{dust3r_cvpr24,
      title={DUSt3R: Geometric 3D Vision Made Easy}, 
      author={Shuzhe Wang and Vincent Leroy and Yohann Cabon and Boris Chidlovskii and Jerome Revaud},
      booktitle = {CVPR},
      year = {2024}
}