remesh-tools-bin

remesh-tools-bin is a Python packaging wrapper for native remeshing command line tools. It currently packages the official valette/ACVD remeshing tools and a small OpenVDB SDF remeshing tool.

Installation

Install from PyPI

uv add remesh-tools-bin

Prebuilt wheels support CPython 3.10-3.13 on Linux x86_64 or ARM64, Windows x86_64, and macOS 12 or newer on Intel or Apple Silicon.

Headless remeshing requires no additional OS packages beyond the platform's baseline runtime libraries. The optional ACVD --display modes still require a working system display and graphics stack.

Build from Source

git clone https://github.com/JYChen18/remesh-tools-bin.git
cd remesh-tools-bin
uv sync
uv build --wheel

The wheel is written to dist/.

Usage

ACVD Remeshing

ACVD methods remesh a surface to a target vertex count. --gradation 0 requests uniform remeshing; higher values give more influence to local curvature. A gradation of 1.5 with manifold output is a good default:

uv run remesh acvd input.obj output.ply --vertices 3000 --gradation 1.5 --force-manifold 1

Available ACVD methods:

• acvd
• acvd-parallel
• acvd-quadric
• acvd-quadric-parallel
• acvd-anisotropic
• acvd-anisotropic-quadric
• acvd-anisotropic-quadric-parallel

Common options include --vertices, --gradation, --subsample, --split-long-edges, and --display. Parallel methods also accept --threads.

OpenVDB SDF Remeshing

The OpenVDB method reads OBJ meshes and writes OBJ meshes by converting the input surface to a signed distance field and extracting a new surface:

uv run remesh openvdb-sdf input.obj output.obj --resolution 50 --level-set 0.1

By default it normalizes the input mesh before applying the OpenVDB conversion and then recovers the original scale, matching the way CoACD uses this preprocessing step. Use --no-normalize to apply OpenVDB directly in input coordinates.

Acknowledgements

The ACVD tools are provided by the vendored valette/ACVD source at commit 275554980e466914ae9053c8667006f251989422. ACVD credits CNRS, INSA-Lyon, UCBL, and INSERM, and derives from the following work:

• S. Valette, J.-M. Chassery, and R. Prost, "Generic remeshing of 3D triangular meshes with metric-dependent discrete Voronoi Diagrams," IEEE TVCG 14(2), 369-381, 2008.
• S. Valette and J.-M. Chassery, "Approximated Centroidal Voronoi Diagrams for Uniform Polygonal Mesh Coarsening," Computer Graphics Forum 23(3), 381-389, 2004.
• M. Audette et al., "Approach-guided controlled resolution brain meshing for FE-based interactive neurosurgery simulation," MICCAI Mesh Processing in Medical Image Analysis Workshop, 2011.

The package also uses OpenVDB for signed-distance-field processing, oneTBB for threading, Boost components required by OpenVDB, and VTK for native mesh processing and Python runtime support.

Third-Party Software

License

Code authored for this repository is licensed under the Apache License 2.0. Bundled third-party components remain under their respective licenses listed above.