pypoisson2 1.0.0

Modern Python bindings for Poisson Surface Reconstruction (v18.74)

pypoisson2

A new Python bindings for Poisson Reconstruction with optional grid output, density values, and gradient vectors.

The corresponding C++ version: v18.74 of PoissonRecon by mkazhdan.

Features

• Poisson surface reconstruction via C++ libraries (int32/int64 auto-selection)
• Optional implicit grid field output
• Boundary condition support (neumann/dirichlet/free)
• Optional density and gradient outputs

Requirements

• Python 3.9+
• NumPy
• Compiled PoissonRecon shared libraries in poisson/lib/, lib/, or build/
• C++ build deps: cmake, g++, make, zlib, libpng, libjpeg, OpenMP

Installation

pip install -e .

Quick Start

# Build shared libs
cmake -S . -B build -DCMAKE_BUILD_TYPE=Release
make -C build -j$(nproc)

# Install Python package
pip install -e .

# Run a basic reconstruction
python - <<'PY'
import numpy as np
from poisson import poisson_reconstruction

points = np.random.randn(1000, 3).astype(np.float64)
points /= np.linalg.norm(points, axis=1, keepdims=True)
normals = points.copy()

vertices, faces = poisson_reconstruction(points, normals, depth=6)
print(vertices.shape, faces.shape)
PY

Build (PoissonRecon + Dump Tool)

This repo builds two executables:

• PoissonRecon in build/
• PoissonReconDump in build_dump/ (exports density/gradients for test alignment)

cmake -S . -B build -DCMAKE_BUILD_TYPE=Release
make -C build -j$(nproc)

cmake -S poisson_dump -B build_dump -DCMAKE_BUILD_TYPE=Release
make -C build_dump -j$(nproc)

Note: Do not define FAST_COMPILE if you need full boundary/degree support.

PyPI Release (Maintainers)

python -m pip install --upgrade build twine
python -m build
python -m twine check dist/*
python -m twine upload dist/*

Basic Usage

import numpy as np
from poisson import poisson_reconstruction

points = np.random.randn(1000, 3).astype(np.float64)
points /= np.linalg.norm(points, axis=1, keepdims=True)

normals = points.copy()

vertices, faces = poisson_reconstruction(points, normals, depth=6)

Advanced Outputs

vertices, faces, grid, iso_value, densities, gradients = poisson_reconstruction(
    points,
    normals,
    depth=6,
    grid_depth=5,
    output_density=True,
    output_gradients=True,
)

Core Parameters

poisson_reconstruction exposes core solver parameters:

• depth, full_depth, cg_depth, iters, degree
• scale, samples_per_node, point_weight, confidence
• exact_interpolation, show_residual, low_depth_cutoff
• width, cg_solver_accuracy, base_depth, solve_depth, kernel_depth, base_v_cycles
• parallel_type, grid_depth
• validate_finite (default True), force_big (force int64 library)

Tests

Tests use unittest and are run as modules (not pytest).

python -m unittest poisson.tests.exe.test_exe_consistency_core
python -m unittest poisson.tests.exe.test_exe_consistency_params
python -m unittest poisson.tests.exe.test_exe_boundary
python -m unittest poisson.tests.exe.test_dump_outputs
python -m unittest poisson.tests.python.test_python_validation

Executable paths (optional overrides):

• POISSON_EXE_PATH defaults to build/PoissonRecon
• POISSON_DUMP_EXE_PATH defaults to build_dump/PoissonReconDump

Notes:

• Boundary tests default to the horse dataset: poisson/examples/horse_with_normals.xyz.
• PoissonReconDump supports ASCII PLY inputs (x y z nx ny nz).