IMODAL¶
Implicit Modular Deformations Analysis Library¶
IMODAL is a python librairy allowing to register shapes (curves, meshes, images) with structured large deformations. The structures are incorporated via deformation modules which generate vector fields of particular, chosen types. They can be defined explicitly (generating local scalings or rotations for instance) or implicitly from constraints. In addition, it is possible to combine them so that a complex structure can be easily defined as the superimposition of simple ones. Trajectories of such modular vector fields can then be integrated to build modular large deformations. Their parameters can be optimized to register observed shapes and analyzed.
Here is an example of reconstruction of basipetal growth using IMODAL:
![_images/basipetal_learntmodel_reshoot_deformed_growth_0.png](_images/basipetal_learntmodel_reshoot_deformed_growth_0.png)
![_images/basipetal_learntmodel_reshoot_deformed_growth_3.png](_images/basipetal_learntmodel_reshoot_deformed_growth_3.png)
![_images/basipetal_learntmodel_reshoot_deformed_growth_5.png](_images/basipetal_learntmodel_reshoot_deformed_growth_5.png)
![_images/basipetal_learntmodel_reshoot_deformed_growth_8.png](_images/basipetal_learntmodel_reshoot_deformed_growth_8.png)
![_images/basipetal_learntmodel_reshoot_deformed_growth_10.png](_images/basipetal_learntmodel_reshoot_deformed_growth_10.png)
IMODAL provides:
Registration of points clouds, curves, meshes and images
Atlas computation with hypertemplate
Estimation of the model parameters
tools to speed up and reduce the memory footprint ( such as GPU and KeOps support)
Authors:
Benjamin Charlier
Barbara Gris
Leander Lacroix
Alain Trouvé
Related publications:
A sub-Riemannian modular framework for diffeomorphism based analysis of shape ensembles, B. Gris, S. Durrleman and A. Trouvé, SIAM Journal of Imaging Sciences, 2018.
IMODAL: creating learnable user-defined deformation models, B. Charlier, L. Lacroix, B. Gris, A. Trouvé, CVPR, 2021.
The project can be downloaded here.
Table of content¶
Understanding IMODAL
Using IMODAL