mrOpenRNb

March 10, 2026 · View on GitHub

mrOpenRNb

Meshroom Plugin for Open-RNb

Integrate Open-RNb neural surface reconstruction directly into your Meshroom photogrammetry pipeline.

Open-RNb is an open-source method for high-quality 3D surface reconstruction from multi-view normal and reflectance (albedo) maps, estimated by photometric stereo methods such as SDM-UniPS and Uni-MS-PS. Built on NeuS neural implicit surfaces and instant-nsr-pl, it combines normal supervision with a two-phase albedo scaling pipeline to produce accurate geometry even when per-view reflectance maps have inconsistent scales.

Brief history

RNb-NeuS (CVPR 2024): the original method introducing reflectance and normal-based multi-view reconstruction. Project page | Paper
RNb-NeuS2 (IJCV 2025): an extended journal version that represents reflectance and surface normals as radiance vectors under simulated illumination, enabling integration into both traditional multi-view stereo and neural volume rendering pipelines. Achieves state-of-the-art results on DiLiGenT-MV, LUCES-MV, and Skoltech3D benchmarks. Project page | arXiv
Open-RNb: a fully open-source reimplementation equivalent to RNb-NeuS2, replacing all proprietary CUDA libraries with standard PyTorch + tiny-cuda-nn. This is the implementation used by this plugin. GitHub

Requirements

Python 3.10+
CUDA 12.x + NVIDIA GPU (RTX 2080 Ti or newer recommended)
Meshroom 2025+ (develop branch)
tinycudann
nerfacc 0.3.3
torch_efficient_distloss

Full dependency list: requirements.txt

Quick Start

Prerequisite: a working Meshroom installation.

1. Clone the plugin

cd /path/to/your/plugins
git clone https://github.com/meshroomHub/mrOpenRNb.git

2. Clone the Open-RNb core code

git clone https://github.com/RobinBruneau/Open-RNb.git

3. Set up the virtual environment

Meshroom looks for a folder named venv at the plugin root and uses its Python interpreter to run the node. You have two options:

Option A: Symlink an existing venv

If you already have a working virtual environment from the Open-RNb repository (e.g. its .venv), you can simply symlink it:

cd mrOpenRNb
ln -s /absolute/path/to/Open-RNb/.venv venv

Option B: Create a fresh venv

cd mrOpenRNb

# Create the venv (must be named "venv", not ".venv")
python3 -m venv venv
source venv/bin/activate

# Upgrade pip and install build tools
pip install --upgrade pip
pip install setuptools==69.5.1 ninja

# Install PyTorch (CUDA 12.x)
pip install torch torchvision

# Install main dependencies
pip install -r requirements.txt

# Install tinycudann
pip install "git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch"

# If the pip install above fails (ModuleNotFoundError: pkg_resources), build from source:
cd /tmp
git clone --recursive https://github.com/NVlabs/tiny-cuda-nn.git
cd tiny-cuda-nn/bindings/torch
# Set your GPU architecture: 70=V100, 75=T4, 80=A100, 86=RTX 3080/3090, 89=RTX 4090
TCNN_CUDA_ARCHITECTURES=86 python setup.py install
cd /path/to/your/plugins/mrOpenRNb

deactivate

4. Configure the plugin

Edit meshroom/config.json to point to your Open-RNb clone:

[
    {
        "key": "OPEN_RNB_PATH",
        "type": "path",
        "value": "/absolute/path/to/Open-RNb"
    }
]

5. Register the plugin in Meshroom

Set the MESHROOM_PLUGINS_PATH environment variable:

# Linux
export MESHROOM_PLUGINS_PATH=/path/to/your/plugins/mrOpenRNb:$MESHROOM_PLUGINS_PATH

# Windows
set MESHROOM_PLUGINS_PATH=C:\path\to\mrOpenRNb;%MESHROOM_PLUGINS_PATH%

Launch Meshroom: the OpenRNb node appears under the Neural Reconstruction category.

6. Verify installation

source venv/bin/activate
python -c "
import tinycudann; print('tinycudann OK')
import nerfacc; print('nerfacc OK')
from torch_efficient_distloss import flatten_eff_distloss; print('distloss OK')
import torch; print(f'CUDA: {torch.cuda.is_available()}, GPU: {torch.cuda.get_device_name(0)}')
"

Plugin Structure

mrOpenRNb/
├── meshroom/
│   ├── config.json                # Plugin configuration (OPEN_RNB_PATH)
│   └── OpenRNb/
│       ├── __init__.py
│       └── OpenRNb.py             # Meshroom node definition
├── venv/                          # Python virtual environment (or symlink, see step 3)
├── requirements.txt               # Python dependencies
└── README.md

For more details on how Meshroom plugins work, see:

Meshroom Plugin Install Guide
mrHelloWorld: step-by-step tutorials for building Meshroom plugins

Node Parameters

Inputs

Parameter	Label	Description
`inputNormalSfm`	Normal Maps SfMData	SfMData file pointing to normal map images (required)
`inputAlbedoSfm`	Albedo Maps SfMData	SfMData file pointing to albedo images (enables two-phase training)
`inputMaskSfm`	Mask SfMData	SfMData file pointing to mask images
`inputMaskFolder`	Mask Folder	Folder with mask images containing viewId in filename (auto-generates SfMData)
`maxSteps`	Max Training Steps	Total training iterations (default: 20000)
`meshResolution`	Mesh Resolution	Marching cubes grid resolution (default: 512)
`scalingMode`	Scaling Mode	Scene normalization: `auto` (prefers silhouettes when masks available, then pcd, then cameras), `pcd`, `silhouettes`, `cameras`, `none`
`sphereScale`	Sphere Scale	Bounding sphere scale after normalization (default: 1.0)
`warmupRatio`	Phase 1 Ratio	Fraction of steps for geometry-only phase (default: 0.1)
`useGpu`	Use GPU	Use GPU for training (default: true)
`openRnbPath`	Open-RNb Path	Path to Open-RNb code (set via `config.json`)

Outputs

Parameter	Description
`outputFolder`	Folder containing all training artifacts
`outputMesh`	Reconstructed OBJ mesh in world coordinates

Acknowledgements

This work is supported by DOPAMIn (Diffusion Open de Photogrammetrie par AliceVision/Meshroom pour l'Industrie), selected in the 2024 cohort of the OPEN programme run by CNRS Innovation. OPEN supports the valorization of open-source scientific software by providing dedicated developer resources, governance expertise, and industry partnership support.

Lead researcher: Jean-Denis Durou, IRIT (INP-Toulouse) Co-lead: Lilian Calvet, Balgrist University Hospital

Project	Description
Open-RNb	Open-source reimplementation of RNb-NeuS2, used by this plugin
RNb-NeuS2	Original RNb-NeuS2 method (IJCV 2025)
mrSDMUniPS	Meshroom plugin for SDM-UniPS photometric stereo
mrLINOUniPS	Meshroom plugin for LINO_UniPS photometric stereo
mrUniMSPS	Meshroom plugin for Uni-MS-PS photometric stereo

Citation

If you use this work, please cite:

@article{bruneau25,
    title={{Multi-view Surface Reconstruction Using Normal and Reflectance Cues}},
    author={Robin Bruneau and Baptiste Brument and Yvain Qu{\'e}au and Jean M{\'e}lou
            and Fran{\c{c}}ois Bernard Lauze and Jean-Denis Durou and Lilian Calvet},
    journal={International Journal of Computer Vision (IJCV)},
    year={2025},
    eprint={2506.04115},
    archivePrefix={arXiv},
    primaryClass={cs.CV},
    url={https://arxiv.org/abs/2506.04115},
}

@inproceedings{brument24,
    title={{RNb-NeuS: Reflectance and Normal-based Multi-View 3D Reconstruction}},
    author={Baptiste Brument and Robin Bruneau and Yvain Qu{\'e}au and Jean M{\'e}lou
            and Fran{\c{c}}ois Lauze and Jean-Denis Durou and Lilian Calvet},
    booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision
               and Pattern Recognition},
    year={2024}
}

License

This project is licensed under the Mozilla Public License 2.0.