Learning the geometry of wave-based imaging

Kothari, Konik and de Hoop, Maarten and Dokmanić, Ivan. (2020) Learning the geometry of wave-based imaging. In: Advances in Neural Information Processing Systems 33 (NeurIPS 2020). pp. 1-12.

Full text not available from this repository.

Official URL: https://edoc.unibas.ch/81812/

Downloads: Statistics Overview


We propose a general physics-based deep learning architecture for wave-based imaging problems. A key difficulty in imaging problems with a varying background wave speed is that the medium “bends” the waves differently depending on their position and direction. This space-bending geometry makes the equivariance to translations of convolutional networks an undesired inductive bias. We build an interpretable neural architecture inspired by Fourier integral operators (FIOs) which approximate the wave physics. FIOs model a wide range of imaging modalities, from seismology and radar to Doppler and ultrasound. We focus on learning the geometry of wave propagation captured by FIOs, which is implicit in the data, via a loss based on optimal transport. The proposed FIONet performs significantly better than the usual baselines on a number of imaging inverse problems, especially in out-of-distribution tests.
Faculties and Departments:05 Faculty of Science > Departement Mathematik und Informatik > Informatik > Signal and Data Analytics (Dokmanic)
UniBasel Contributors:Dokmanić, Ivan
Item Type:Conference or Workshop Item, refereed
Conference or workshop item Subtype:Conference Paper
Publisher:Curran Associates, Inc
Note:Publication type according to Uni Basel Research Database: Conference paper
Last Modified:27 Jul 2021 13:24
Deposited On:27 Jul 2021 13:24

Repository Staff Only: item control page