fast-rs 1.0.0

Fast sphericity and roundness in 2D and 3D using local thickness

Fast sphericity and roundness approximation in 2D and 3D

Fast approximation of Wadell sphericity and roundness in 2D and 3D, described in our paper (link), (arXiv).

The method uses the output from local thickness algorithm to approximate the two measures. Resulting sphericity values closely match those from exact approaches. Roundness values no longer retain the original 0-1 range, but correlate closely to it.

The execution speed is dependent primarily on image volume, and not on number of objects, enabling fast evaluation of thousands of objects at once.

For in-depth introduction to the library, visit the CodeOcean capsule.

Installation

pip install fast_rs

How to use

For in-depth introduction, visit the CodeOcean capsule.

Both sphericity and roundness can be calculated together, from a binary mask. The process is exactly the same in 2D and 3D.

from fast_rs import rs

mask = ... # Load/provide binary mask (can be both a 2D and 3D np.array)
roundness_vec, sphericity_vec, label_img = rs.rs_calculate(mask)

Returned label image indices can be used to connect measure values to specific objects in the mask.

Example result:

Paper

The fast sphericity and roundness approximation method is described and evaluated in our contribution to CVMI (CVPR 2025 Workshop). Please cite our paper if you use the method in your work.

@article{pieta2025a,
  author = {Pieta, Pawel Tomasz and Rasmussen, Peter Winkel and Dahl, Anders Bjorholm and Christensen, Anders Nymark},
  title = {Fast Sphericity and Roundness Approximation in 2D and 3D Using Local Thickness},
  language = {eng},
  format = {article},
  journal = {Proceedings of 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)},
  pages = {4667-4677},
  year = {2025},
  publisher = {IEEE},
  doi = {10.1109/CVPRW67362.2025.00453}
}

License

MIT License (see LICENSE file).