DeepInverse: A Python package for solving imaging inverse problems with deep learning

TL;DR

DeepInverse is an open-source Python library that simplifies solving imaging inverse problems using deep learning, covering forward operators, variational problems, and neural network training.

Contribution

It introduces a comprehensive, PyTorch-based toolkit for image reconstruction tasks, integrating multiple steps in the inverse problem-solving pipeline.

Findings

Efficient implementation of forward operators for various imaging modalities

Flexible framework for defining and solving variational problems

Support for designing and training advanced neural network architectures

Abstract

DeepInverse is an open-source PyTorch-based library for solving imaging inverse problems. The library covers all crucial steps in image reconstruction from the efficient implementation of forward operators (e.g., optics, MRI, tomography), to the definition and resolution of variational problems and the design and training of advanced neural network architectures. In this paper, we describe the main functionality of the library and discuss the main design choices.