Filters








877 Hits in 2.9 sec

Algorithmic Guarantees for Inverse Imaging with Untrained Network Priors [article]

Gauri Jagatap, Chinmay Hegde
2020 arXiv   pre-print
However, few theoretical guarantees exist in the scope of using untrained neural network priors for inverse imaging problems.  ...  We model images to lie in the range of an untrained deep generative network with a fixed seed.  ...  We further provide preliminary guarantees for the convergence of a projected gradient descent scheme to solve the problem of compressive phase retrieval.  ... 
arXiv:1906.08763v2 fatcat:lntgt2cqh5d35eayht5mah467e

Low Shot Learning with Untrained Neural Networks for Imaging Inverse Problems [article]

Oscar Leong, Wesam Sakla
2019 arXiv   pre-print
Employing deep neural networks as natural image priors to solve inverse problems either requires large amounts of data to sufficiently train expressive generative models or can succeed with no data via  ...  untrained neural networks.  ...  The authors use an untrained neural network as a natural image prior and, when given a small amount of data, adopt a learned regularization term when solving the inverse problem.  ... 
arXiv:1910.10797v1 fatcat:sjik45u5izctbddhzscxh4sz6a

On Architecture Selection for Linear Inverse Problems with Untrained Neural Networks

Yang Sun, Hangdong Zhao, Jonathan Scarlett
2021 Entropy  
In recent years, neural network based image priors have been shown to be highly effective for linear inverse problems, often significantly outperforming conventional methods that are based on sparsity  ...  While pre-trained generative models are perhaps the most common, it has additionally been shown that even untrained neural networks can serve as excellent priors in various imaging applications.  ...  As an alternative method that can help overcome the above limitations, it has been observed that even untrained neural networks can serve as excellent priors for image recovery in linear inverse problems  ... 
doi:10.3390/e23111481 pmid:34828179 pmcid:PMC8623203 fatcat:qna2tmzjjvdcfiigcnpfwtlena

Untrained networks for compressive lensless photography [article]

Kristina Monakhova, Vi Tran, Grace Kuo, Laura Waller
2021 arXiv   pre-print
In this work, we propose the use of untrained networks for compressive image recovery.  ...  They have been demonstrated for 2D and 3D microscopy, single-shot video, and single-shot hyperspectral imaging; in each of these cases, a compressive-sensing-based inverse problem is solved in order to  ...  We believe that untrained networks are especially promising for situations in which training data is difficult or impossible to obtain, providing a better imaging prior for underdetermined reconstructions  ... 
arXiv:2103.07609v1 fatcat:3kcpcevxq5bj7n2vxa4qwy4mxu

A Consensus Equilibrium Solution For Deep Image Prior Powered By Red

Rakib Hyder, Hassan Mansour, Yanting Ma, Petros T. Boufounos, Pu Wang
2021 ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)  
Recent advances in solving imaging inverse problems have witnessed the combination of deep learning models with classical image models for better signal representation.  ...  Index Terms-Inverse problem, image deblurring, consensus equilibrium, deep image prior, RED  ...  In another line of work, untrained convolutional network architectures have been used as image prior.  ... 
doi:10.1109/icassp39728.2021.9414290 fatcat:277pc7q6jveclc5hqpa3eoyx2y

Generative Patch Priors for Practical Compressive Image Recovery [article]

Rushil Anirudh, Suhas Lohit, Pavan Turaga
2020 arXiv   pre-print
In this paper, we propose the generative patch prior (GPP) that defines a generative prior for compressive image recovery, based on patch-manifold models.  ...  Unlike learned, image-level priors that are restricted to the range space of a pre-trained generator, GPP can recover a wide variety of natural images using a pre-trained patch generator.  ...  Algorithmic guaran- tees for inverse imaging with untrained network priors. 2019. [27] R. Kerviche, N. Zhu, and A. Ashok. Information- optimal scalable compressive imaging system.  ... 
arXiv:2006.10873v2 fatcat:kn6b6n4csfhwnoy6bu5uhsyuua

JR2net: A Joint Non-Linear Representation and Recovery Network for Compressive Spectral Imaging [article]

Brayan Monroy, Jorge Bacca, Henry Arguello
2022 arXiv   pre-print
For instance, the deep spectral prior approach uses a convolutional autoencoder network (CAE) in the optimization algorithm to recover the spectral image by using a non-linear representation.  ...  However, the CAE training is detached from the recovery problem, which does not guarantee optimal representation of the spectral images for the CSI problem.  ...  Model-based Optimization with Deep Priors These methods are based on iterative techniques that replace the hand-crafted prior with a deep neural network (DNN) used to learn a deep prior of the spectral  ... 
arXiv:2205.07770v1 fatcat:sajbzktaz5akjihfrlfxlrq7se

A Plug-and-Play Deep Image Prior

Zhaodong Sun, Fabian Latorre, Thomas Sanchez, Volkan Cevher
2021 ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)  
Deep image priors (DIP) offer a novel approach for the regularization that leverages the inductive bias of a deep convolutional architecture in inverse problems.  ...  To mitigate this effect, this work incorporates a plug-andplay prior scheme which can accommodate additional regularization steps within a DIP framework.  ...  Deep Image Prior (DIP) [7] is a different approach, conceptually close to CSGM [6] but different in a crucial way: it leverages the inductive bias of an untrained convolutional neural network architecture  ... 
doi:10.1109/icassp39728.2021.9414879 fatcat:rhywqaeb6nevjc5cvpwee3vjgy

A Unifying Multi-sampling-ratio CS-MRI Framework With Two-grid-cycle Correction and Geometric Prior Distillation [article]

Xiaohong Fan, Yin Yang, Ke Chen, Jianping Zhang, Ke Dong
2022 arXiv   pre-print
and geometric prior distillation module.  ...  guarantee of model-based methods and the superior reconstruction performances of deep learning-based methods.  ...  However, if we incorporate the learnable geometric prior distillation stage (P2) into the solution fidelity of image inverse problem (1), the network can learn the expected MR image priors.  ... 
arXiv:2205.07062v1 fatcat:xhnnu2zatjepnjaw6xv6twt2qa

Deep Learning Techniques for Inverse Problems in Imaging [article]

Gregory Ongie, Ajil Jalal, Christopher A. Metzler, Richard G. Baraniuk, Alexandros G. Dimakis, Rebecca Willett
2020 arXiv   pre-print
Recent work in machine learning shows that deep neural networks can be used to solve a wide variety of inverse problems arising in computational imaging.  ...  We also discuss the trade-offs associated with these different reconstruction approaches, caveats and common failure modes, plus open problems and avenues for future work.  ...  Acknowledgments We thank Davis Gilton for performing the MRI reconstruction experiments pictured in  ... 
arXiv:2005.06001v1 fatcat:z7w3vygugjf57fqbe6t62fvni4

Limited-view Photoacoustic Imaging Reconstruction With Dual Domain Inputs Under Mutual Information Constraint [article]

Jiadong Zhang, Hengrong Lan, Changchun Yang, Shanshan Guo, Feng Gao, Fei Gao
2020 arXiv   pre-print
Besides, we use mutual information (MI) with an auxiliary network, whose inputs and outputs are both ground truth, to compensate prior knowledge of limited-view PA inputs.  ...  With conventional image reconstruction algorithms, the limited-view tissue induces artifacts and information loss, which may cause doctors misdiagnosis or missed diagnosis.  ...  Because auxiliary network is pre-trained, z 1 is untrainable.  ... 
arXiv:2011.06147v1 fatcat:lez3bguhq5adljoh2n5afxc6ce

Compressed Sensing with Deep Image Prior and Learned Regularization [article]

Dave Van Veen, Ajil Jalal, Mahdi Soltanolkotabi, Eric Price, Sriram Vishwanath, Alexandros G. Dimakis
2020 arXiv   pre-print
Our method is based on the recently proposed Deep Image Prior (DIP), wherein the convolutional weights of the network are optimized to match the observed measurements.  ...  We propose a novel method for compressed sensing recovery using untrained deep generative models.  ...  [79] proposed Deep Image Prior (DIP), which uses untrained convolutional neural networks. In DIP-based schemes, a convolutional neural network generator (e.g.  ... 
arXiv:1806.06438v4 fatcat:hqjraly4vrhwbnhpjvsci2fosa

Inverting Gradients – How easy is it to break privacy in federated learning? [article]

Jonas Geiping, Hartmut Bauermeister, Hannah Dröge, Michael Moeller
2020 arXiv   pre-print
Previous attacks have provided a false sense of security, by succeeding only in contrived settings - even for a single image.  ...  from the knowledge of their parameter gradients, and demonstrate that such a break of privacy is possible even for trained deep networks.  ...  not interfere with the collaborative learning algorithm.  ... 
arXiv:2003.14053v2 fatcat:cuzoihte3vfnffzzl7wivjttnq

Gradient Inversion with Generative Image Prior [article]

Jinwoo Jeon and Jaechang Kim and Kangwook Lee and Sewoong Oh and Jungseul Ok
2021 arXiv   pre-print
Further, when such prior knowledge is unavailable, we investigate the possibility of learning the prior from a sequence of gradients seen in the process of FL training.  ...  However, a gradient is often insufficient to reconstruct the user data without any prior knowledge.  ...  Our experiment code is available at https://github.com/ml-postech/ gradient-inversion-generative-image-prior. Algorithms. We evaluate several algorithms for the gradient inversion (GI) task in (3).  ... 
arXiv:2110.14962v1 fatcat:37elalcgkja4xds42yekvr3cyi

Class-Specific Blind Deconvolutional Phase Retrieval Under a Generative Prior [article]

Fahad Shamshad, Ali Ahmed
2020 arXiv   pre-print
We propose to solve this inverse problem using alternating gradient descent algorithm under two pretrained deep generative networks as priors; one is trained on sharp images and the other on blur kernels  ...  The proposed recovery algorithm strives to find a sharp image and a blur kernel in the range of the respective pre-generators that best explain the forward measurement model.  ...  In [2] , authors aim to solve the phaseless blind image deblurring problem via convex program with rigorous theoretical guarantees.  ... 
arXiv:2002.12578v1 fatcat:p3ioca5zxzfe3o5zdfskc3thvy
« Previous Showing results 1 — 15 out of 877 results