nagini3D 0.2.3

Implementation of NAGINI-3D, a python package designed for Multi-Object Segmentation in Biological Imaging based on deep learning and active surfaces.

NAGINI-3D | Prediction of Parametric Surfaces for Multi-Object Segmentation in 3D Biological Imaging

We present NAGINI-3D (N-Active shapes for seGmentINg 3D biological Images), a method dedicated to 3D biological images segmentation, based on both deep learning (CNN) and Active Surfaces (Snakes).

This package implement the method described in:

• Quentin RAPILLY, Anaïs BADOUAL,Pierre MAINDRON, Guenaelle BOUET, Charles KERVRANN. Prediction of Parametric Surfaces for Multi-Object Segmentation in 3D Biological Imaging. Scale Space and Variational Methods in Computer Vision. SSVM 2025. Lecture Notes in Computer Science, vol 15667, Devon, UK, May 2025, (preprint), (final paper).

All details on the implementation and the tutorials are available on the github of the project.

Versions

0.2.3

Bug fix for the local curvature computation.

0.2.2

New surface regularization strategy that permits to train the network to predict highly parametrized surfaces.

0.1.2

Minor bugs fix.

0.1.1

Anisotropic version: option to process highly anisotropic images.

0.0.2

Paper version: code used to assess our method for SSVM submission.

0.0.1

Test version: first pypi version.