One Network to Solve Them All --- Solving Linear Inverse Problems using Deep Projection Models

Abstract

While deep learning methods have achieved state-of-the-art performance in many challenging inverse problems like image inpainting and super-resolution, they invariably involve problem-specific training of the networks. Under this approach, different problems require different networks. In scenarios where we need to solve a wide variety of problems, e.g., on a mobile camera, it is inefficient and costly to use these specially-trained networks. On the other hand, traditional methods using signal priors can be used in all linear inverse problems but often have worse performance on challenging tasks. In this work, we provide a middle ground between the two kinds of methods --- we propose a general framework to train a single deep neural network that solves arbitrary linear inverse problems. The proposed network acts as a proximal operator for an optimization algorithm and projects non-image signals onto the set of natural images defined by the decision boundary of a classifier. In our experiments, the proposed framework demonstrates superior performance over traditional methods using a wavelet sparsity prior and achieves comparable performance of specially-trained networks on tasks including compressive sensing and pixel-wise inpainting.

Cited by in corpus (28)

• On instabilities of deep learning in image reconstruction - Does AI come at a cost?
• fastMRI: An Open Dataset and Benchmarks for Accelerated MRI
• Compressed Sensing with Deep Image Prior and Learned Regularization
• Learning to Optimize: A Primer and A Benchmark
• Convolutional Proximal Neural Networks and Plug-and-Play Algorithms
• A survey and an extensive evaluation of popular audio declipping methods
• Robust Conditional Generative Adversarial Networks
• prDeep: Robust Phase Retrieval with a Flexible Deep Network
• ADCME: Learning Spatially-varying Physical Fields using Deep Neural Networks
• Improving Robustness of Deep-Learning-Based Image Reconstruction
• Solving Linear Inverse Problems Using GAN Priors: An Algorithm with Provable Guarantees
• Tuning-free Plug-and-Play Proximal Algorithm for Inverse Imaging Problems
• Total Deep Variation: A Stable Regularizer for Inverse Problems
• Deep Learning Techniques for Inverse Problems in Imaging
• Variational Inference for Computational Imaging Inverse Problems
• ISTA-Net: Interpretable Optimization-Inspired Deep Network for Image Compressive Sensing
• Identifying Recurring Patterns with Deep Neural Networks for Natural Image Denoising
• Random mesh projectors for inverse problems
• Subspace Embeddings Under Nonlinear Transformations
• Snapshot Interferometric 3D Imaging by Compressive Sensing and Deep Learning
• Parallel and Flexible Sampling from Autoregressive Models via Langevin Dynamics
• Deep Decomposition Learning for Inverse Imaging Problems
• Scalable Deep Compressive Sensing
• Learned Interpretable Residual Extragradient ISTA for Sparse Coding
• Learning to Learn Graph Topologies
• Improving Model Compatibility of Generative Adversarial Networks by Boundary Calibration
• A Deep Optimization Approach for Image Deconvolution
• Extra Proximal-Gradient Inspired Non-local Network