dival.reconstructors.dip_ct_reconstructor module

class dival.reconstructors.dip_ct_reconstructor.DeepImagePriorCTReconstructor(ray_trafo, callback_func=None, callback_func_interval=100, show_pbar=True, torch_manual_seed=10, **kwargs)

Bases: dival.reconstructors.reconstructor.IterativeReconstructor

CT reconstructor applying DIP with TV regularization (see 2). The DIP was introduced in 1.

References

1

V. Lempitsky, A. Vedaldi, and D. Ulyanov, 2018, “Deep Image Prior”. IEEE/CVF Conference on Computer Vision and Pattern Recognition. https://doi.org/10.1109/CVPR.2018.00984

2

D. Otero Baguer, J. Leuschner, M. Schmidt, 2020, “Computed Tomography Reconstruction Using Deep Image Prior and Learned Reconstruction Methods”. Inverse Problems. https://doi.org/10.1088/1361-6420/aba415

HYPER_PARAMS = {'channels': {'default': [128, 128, 128, 128, 128]}, 'gamma': {'default': 0.0001, 'grid_search_options': {'num_samples': 20}, 'range': [1e-07, 1.0]}, 'iterations': {'default': 5000, 'range': [1, 50000]}, 'loss_function': {'choices': ['mse', 'poisson'], 'default': 'mse'}, 'lr': {'default': 0.001, 'range': [1e-05, 0.1]}, 'mu_max': {'default': 81.35858, 'range': [1.0, 10000.0]}, 'photons_per_pixel': {'default': 4096, 'range': [1000, 10000]}, 'scales': {'choices': [3, 4, 5, 6, 7], 'default': 4}, 'skip_channels': {'default': [4, 4, 4, 4, 4]}}

__init__(ray_trafo, callback_func=None, callback_func_interval=100, show_pbar=True, torch_manual_seed=10, **kwargs)

Parameters

• ray_trafo (odl.tomo.operators.RayTransform) – The forward operator

• callback_func (callable, optional) – Callable with signature callback_func(iteration, reconstruction, loss) that is called after every callback_func_interval iterations, starting after the first iteration. It is additionally called after the last iteration. Note that it differs from the inherited IterativeReconstructor.callback (which is also supported) in that the latter is of type odl.solvers.util.callback.Callback, which only receives the reconstruction, such that the loss would have to be recomputed.

• callback_func_interval (int, optional) – Number of iterations between calls to callback_func. Default: 100.

• show_pbar (bool, optional) – Whether to show a tqdm progress bar during reconstruction.

• torch_manual_seed (int, optional) – Fixed seed to set by torch.manual_seed before reconstruction. The default is 10. It can be set to None or False to disable the manual seed.

get_activation(layer_index)

property channels

property gamma

property iterations

property loss_function

property lr

property mu_max

property photons_per_pixel

property scales

property skip_channels