DeepInverse: a Pytorch library for imaging with deep learning

Deep Inverse is a Pytorch based library for solving imaging inverse problems with deep learning. This library provides a large collection of predefined imaging operators (magnetic resonance imaging, computed tomography, compressed sensing, blurring, inpainting, etc.), popular supervised and unsupervised learning losses (noise2x, equivariant imaging, etc.) and unrolled architectures (ADMM, forward-backward, deep equilibrium, etc.).

Contents:

• Models
  • Simple AutoEncoder
  • Vanilla U-Net
• Physics
  • Denoising
  • Deblurring and super-resolution
  • Inpainting
  • Magnetic Resonance Imaging
  • Computed Tomography
  • Compressed Sensing
  • Dehazing
  • Colorization
  • Forward
• Datasets
  • Data Generator
  • Quick MNIST dataloader
• Loss
  • Equivariant Imaging
  • Other Losses
  • Measurement Consistency
  • Measurement Splitting
  • Multioperator Imaging
  • Supervised Learning
  • Stein’s Unbiased Risk Estimator (SURE)
  • Total Variation
• Optim
  • Optimizers
  • Fixed Point
  • Data fidelity
  • Utils
• Unfolded
  • Unfolded
  • Deep Equilibrium
• PnP
• Transforms
  • Rotate
  • Shift

Getting Started

Here quick guide

Lead Developers

Julian Tachella, Dongdong Chen, Samuel Hurault and Matthieu Terris.

Cite Us

Here how to cite us

Indices and tables

• Index

• Module Index

• Search Page