Bruno Lecouat
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Bruno Lecouat has a total of 5 co-authors. Here are the top co-authors by number of shared publications:
- Jean Ponce
Département d'informatique de l'ENS INRIA ENS CNRS PSL University
2 shared publications - Julien Mairal
University Grenoble Alpes, Inria, CNRS, Grenoble INP, LJK, Grenoble, France
2 shared publications Bruno Lecouat Inria
1 shared publicationJean Ponce
PSL University
1 shared publicationJulien Mairal
Inria, Univ. Grenoble Alpes, CNRS, Grenoble INP, LJK
1 shared publication
Latest Publications
- 17 November 2020
- Published by Springer Nature
- p. 238-254
Fully Trainable and Interpretable Non-local Sparse Models for Image Restoration
- 17 November 2020
- Published by Springer Nature
- p. 238-254
Fully Trainable and Interpretable Non-local Sparse Models for Image Restoration
- Open Access
A Flexible Framework for Designing Trainable Priors with Adaptive Smoothing and Game Encoding
We introduce a general framework for designing and training neural network layers whose forward passes can be interpreted as solving non-smooth convex optimization problems, and whose architectures are derived from an optimization algorithm. We focus on convex games, solved by local agents represented by the nodes of a graph and interacting through regularization functions. This approach is appealing for solving imaging problems, as it allows the use of classical image priors within deep models that are trainable end to end. The priors used in this presentation include variants of total variation, Laplacian regularization, bilateral filtering, sparse coding on learned dictionaries, and non-local self similarities. Our models are fully interpretable as well as parameter and data efficient. Our experiments demonstrate their effectiveness on a large diversity of tasks ranging from image denoising and compressed sensing for fMRI to dense stereo matching.
Open AccessA Flexible Framework for Designing Trainable Priors with Adaptive Smoothing and Game Encoding
We introduce a general framework for designing and training neural network layers whose forward passes can be interpreted as solving non-smooth convex optimization problems, and whose architectures are derived from an optimization algorithm. We focus on convex games, solved by local agents represented by the nodes of a graph and interacting through regularization functions. This approach is appealing for solving imaging problems, as it allows the use of classical image priors within deep models that are trainable end to end. The priors used in this presentation include variants of total variation, Laplacian regularization, bilateral filtering, sparse coding on learned dictionaries, and non-local self similarities. Our models are fully interpretable as well as parameter and data efficient. Our experiments demonstrate their effectiveness on a large diversity of tasks ranging from image denoising and compressed sensing for fMRI to dense stereo matching.
- Open Access
Fully Trainable and Interpretable Non-Local Sparse Models for Image Restoration
Non-local self-similarity and sparsity principles have proven to be powerful priors for natural imagemodeling. We propose a novel differentiable relaxation of joint sparsity that exploits both principlesand leads to a general framework for image restoration which is (1) trainable end to end, (2) fullyinterpretable, and (3) much more compact than competing deep learning architectures. We apply thisapproach to denoising, jpeg deblocking, and demosaicking, and show that, with as few as 100K parameters,its performance on several standard benchmarks is on par or better than state-of-the-art methods thatmay have an order of magnitude or more parameters.
Open AccessFully Trainable and Interpretable Non-Local Sparse Models for Image Restoration
Non-local self-similarity and sparsity principles have proven to be powerful priors for natural imagemodeling. We propose a novel differentiable relaxation of joint sparsity that exploits both principlesand leads to a general framework for image restoration which is (1) trainable end to end, (2) fullyinterpretable, and (3) much more compact than competing deep learning architectures. We apply thisapproach to denoising, jpeg deblocking, and demosaicking, and show that, with as few as 100K parameters,its performance on several standard benchmarks is on par or better than state-of-the-art methods thatmay have an order of magnitude or more parameters.