Training Adaptive Reconstruction Networks for Blind Inverse Problems - Signal et Communications Access content directly
Preprints, Working Papers, ... Year : 2023

Training Adaptive Reconstruction Networks for Blind Inverse Problems


Neural networks allow solving many ill-posed inverse problems with unprecedented performance. Physics informed approaches already progressively replace carefully hand-crafted reconstruction algorithms in real applications. However, these networks suffer from a major defect: when trained on a given forward operator, they do not generalize well to a different one. The aim of this paper is twofold. First, we show through various applications that training the network with a family of forward operators allows solving the adaptivity problem without compromising the reconstruction quality significantly.Second, we illustrate that this training procedure allows tackling challenging blind inverse problems.Our experiments include partial Fourier sampling problems arising in magnetic resonance imaging (MRI) with sensitivity estimation and off-resonance effects, computerized tomography (CT) with a tilted geometry and image deblurring with Fresnel diffraction kernels.
Fichier principal
Vignette du fichier
Training_Adaptive_Reconstruction_Networks_for_Blind_Inverse_Problems.pdf (4.42 Mo) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-03585120 , version 1 (22-02-2022)
hal-03585120 , version 2 (25-02-2022)
hal-03585120 , version 3 (13-11-2022)
hal-03585120 , version 4 (13-10-2023)
hal-03585120 , version 5 (13-12-2023)



Alban Gossard, Pierre Weiss. Training Adaptive Reconstruction Networks for Blind Inverse Problems. 2023. ⟨hal-03585120v5⟩
398 View
363 Download



Gmail Mastodon Facebook X LinkedIn More