Registration Model

class Models.RegistrationModel(deformables, deformation_modules, attachments, fit_gd=None, lam=1.0, precompute_callback=None, other_parameters=None)

property attachments

compute_deformed(solver, it, t1=1.0, costs=None, intermediates=None)

Compute the deformed source.

Parameters

• solver (str) – Solver to use for the shooting.

• it (int) – Number of iterations the integration method will do.

• costs (dict, default=None) – If provided, will be filled with the costs associated to the deformation.

Returns

List of deformed sources.

Return type

list

property deformables

property deformation_modules

evaluate(target, solver, it, costs=None, backpropagation=True)

Evaluate the model and output its cost.

Parameters

• targets (torch.Tensor or list of torch.Tensor) – Targets we try to approach.

• solver (str) – Solver to use for the shooting.

• it (int) – Number of iterations for the integration.

Returns

Dictionnary of (string, float) pairs, representing the costs.

Return type

dict

fill_init_manifold(init_manifold)

property fit_gd

property init_manifold

property init_other_parameters

property lam

property modules

property parameters

property precompute_callback

to_device(device)