EmNeF: Neural Fields for Embedded Variational Problems in Imaging

Danielle Bednarski; Jan Lellmann

doi:10.1007/978-3-031-31975-4_11

EmNeF: Neural Fields for Embedded Variational Problems in Imaging

Danielle Bednarski^*, Jan Lellmann

^*Corresponding author for this work

Institute of Mathematics and Image Computing

Abstract

We propose a model-driven neural fields approach for solving variational problems. The approach can be applied to a variety of problems with convex, 1-homogeneous regularizer and arbitrary, possibly non-convex, data term. Our strategy is to embed the non-convex energy into a higher-dimensional space, reaching a convex primal-dual formulation. Instead of using classical gradient-descent based optimization algorithms, we propose training multiple fields representing the primal and dual variables in order to solve the problem.

Original language	English
Title of host publication	Scale Space and Variational Methods in Computer Vision : 9th International Conference, SSVM 2023, Santa Margherita di Pula, Italy, May 21-25, 2023, Proceedings
Number of pages	12
Publisher	Springer, Cham
Publication date	10.05.2023
Pages	137 - 148
ISBN (Print)	9783031319747
DOIs	https://doi.org/10.1007/978-3-031-31975-4_11
Publication status	Published - 10.05.2023

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 9 Industry, Innovation, and Infrastructure

Access to Document

10.1007/978-3-031-31975-4_11

Cite this

Bednarski, D., & Lellmann, J. (2023). EmNeF: Neural Fields for Embedded Variational Problems in Imaging. In Scale Space and Variational Methods in Computer Vision: 9th International Conference, SSVM 2023, Santa Margherita di Pula, Italy, May 21-25, 2023, Proceedings (pp. 137 - 148). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14009 LNCS). Springer, Cham. https://doi.org/10.1007/978-3-031-31975-4_11

Bednarski, Danielle ; Lellmann, Jan. / EmNeF: Neural Fields for Embedded Variational Problems in Imaging. Scale Space and Variational Methods in Computer Vision: 9th International Conference, SSVM 2023, Santa Margherita di Pula, Italy, May 21-25, 2023, Proceedings. Springer, Cham, 2023. pp. 137 - 148 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{d7b545cdbd2a44eb82380c0ad68c9536,

title = "EmNeF: Neural Fields for Embedded Variational Problems in Imaging",

abstract = "We propose a model-driven neural fields approach for solving variational problems. The approach can be applied to a variety of problems with convex, 1-homogeneous regularizer and arbitrary, possibly non-convex, data term. Our strategy is to embed the non-convex energy into a higher-dimensional space, reaching a convex primal-dual formulation. Instead of using classical gradient-descent based optimization algorithms, we propose training multiple fields representing the primal and dual variables in order to solve the problem.",

author = "Danielle Bednarski and Jan Lellmann",

note = "DBLP License: DBLP's bibliographic metadata records provided through http://dblp.org/ are distributed under a Creative Commons CC0 1.0 Universal Public Domain Dedication. Although the bibliographic metadata records are provided consistent with CC0 1.0 Dedication, the content described by the metadata records is not. Content may be subject to copyright, rights of privacy, rights of publicity and other restrictions.",

year = "2023",

month = may,

day = "10",

doi = "10.1007/978-3-031-31975-4\_11",

language = "English",

isbn = "9783031319747",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer, Cham",

pages = " 137 -- 148",

booktitle = "Scale Space and Variational Methods in Computer Vision",

}

Bednarski, D & Lellmann, J 2023, EmNeF: Neural Fields for Embedded Variational Problems in Imaging. in Scale Space and Variational Methods in Computer Vision: 9th International Conference, SSVM 2023, Santa Margherita di Pula, Italy, May 21-25, 2023, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 14009 LNCS, Springer, Cham, pp. 137 - 148. https://doi.org/10.1007/978-3-031-31975-4_11

EmNeF: Neural Fields for Embedded Variational Problems in Imaging. / Bednarski, Danielle; Lellmann, Jan.
Scale Space and Variational Methods in Computer Vision: 9th International Conference, SSVM 2023, Santa Margherita di Pula, Italy, May 21-25, 2023, Proceedings. Springer, Cham, 2023. p. 137 - 148 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14009 LNCS).

Research output: Chapters in Books/Reports/Conference Proceedings › Conference contribution › peer-review

TY - GEN

T1 - EmNeF: Neural Fields for Embedded Variational Problems in Imaging

AU - Bednarski, Danielle

AU - Lellmann, Jan

N1 - DBLP License: DBLP's bibliographic metadata records provided through http://dblp.org/ are distributed under a Creative Commons CC0 1.0 Universal Public Domain Dedication. Although the bibliographic metadata records are provided consistent with CC0 1.0 Dedication, the content described by the metadata records is not. Content may be subject to copyright, rights of privacy, rights of publicity and other restrictions.

PY - 2023/5/10

Y1 - 2023/5/10

N2 - We propose a model-driven neural fields approach for solving variational problems. The approach can be applied to a variety of problems with convex, 1-homogeneous regularizer and arbitrary, possibly non-convex, data term. Our strategy is to embed the non-convex energy into a higher-dimensional space, reaching a convex primal-dual formulation. Instead of using classical gradient-descent based optimization algorithms, we propose training multiple fields representing the primal and dual variables in order to solve the problem.

AB - We propose a model-driven neural fields approach for solving variational problems. The approach can be applied to a variety of problems with convex, 1-homogeneous regularizer and arbitrary, possibly non-convex, data term. Our strategy is to embed the non-convex energy into a higher-dimensional space, reaching a convex primal-dual formulation. Instead of using classical gradient-descent based optimization algorithms, we propose training multiple fields representing the primal and dual variables in order to solve the problem.

UR - https://rdcu.be/dhGx9

UR - https://www.mendeley.com/catalogue/e0fd50ce-8c3d-3050-9de7-1f8bac969cce/

U2 - 10.1007/978-3-031-31975-4_11

DO - 10.1007/978-3-031-31975-4_11

M3 - Conference contribution

SN - 9783031319747

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 137

EP - 148

BT - Scale Space and Variational Methods in Computer Vision

PB - Springer, Cham

ER -

Bednarski D, Lellmann J. EmNeF: Neural Fields for Embedded Variational Problems in Imaging. In Scale Space and Variational Methods in Computer Vision: 9th International Conference, SSVM 2023, Santa Margherita di Pula, Italy, May 21-25, 2023, Proceedings. Springer, Cham. 2023. p. 137 - 148. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-31975-4_11

EmNeF: Neural Fields for Embedded Variational Problems in Imaging

Abstract

UN SDGs

Access to Document

Other files and links

Cite this