pixcurve

High-performance C++ conversion from labeled pixel masks to compact Bezier-style polygons.

Input is a single-channel uint8 label mask where background is 0. Output is JSON-compatible polygons keyed by label value.

Install

python -m pip install pixcurve

For local development from source:

python -m pip install -e .

If a prebuilt wheel is not available for your platform, installation from source requires:

• CMake
• a C++20 compiler
• OpenCV development libraries available to CMake

Python API

import numpy as np
from PIL import Image

from pixcurve import mask_to_bezier

mask = np.asarray(Image.open("mask.png").convert("L"), dtype=np.uint8)
polygons = mask_to_bezier(mask)

Output shape:

{"11":[[[x,y],[x,y]]],"16":[[[x,y],[x,y]]]}

CLI

pixcurve path/to/mask.png > polygons.json

Example

python examples/convert_mask.py path/to/mask.png > polygons.json

Native C++ Binary

cmake -S src_cpp -B src_cpp/build -DCMAKE_BUILD_TYPE=Release
cmake --build src_cpp/build -j
./src_cpp/build/mask_to_bezier_cpp path/to/mask.png

Tests

python -m unittest discover -s tests
cmake -S src_cpp -B src_cpp/build -DCMAKE_BUILD_TYPE=Release
cmake --build src_cpp/build -j
ctest --test-dir src_cpp/build --output-on-failure

Publishing checklist

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

Then upload to TestPyPI first:

python -m twine upload --repository testpypi dist/*