A Primal Dual Network for Low-Level Vision Problems

Christoph Vogel, Thomas Pock

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In the past, classic energy optimization techniques were the driving force in many innovations and are a building block for almost any problem in computer vision. Efficient algorithms are mandatory to achieve real-time processing, needed in many applications like autonomous driving. However, energy models - even if designed by human experts - might never be able to fully capture the complexity of natural scenes and images. Similar to optimization techniques, Deep Learning has changed the landscape of computer vision in recent years and has helped to push the performance of many models to never experienced heights.
Our idea of a primal-dual network is to combine the structure of regular energy optimization techniques, in particular of first order methods, with the flexibility of Deep Learning to adapt to the statistics of the input data.
LanguageGerman
Title of host publicationGerman Conference on Pattern Recognition, 2017
PublisherSpringer Berlin - Heidelberg
StatusPublished - 1 Sep 2017

Cite this

Vogel, C., & Pock, T. (2017). A Primal Dual Network for Low-Level Vision Problems. In German Conference on Pattern Recognition, 2017 Springer Berlin - Heidelberg.

A Primal Dual Network for Low-Level Vision Problems. / Vogel, Christoph; Pock, Thomas.

German Conference on Pattern Recognition, 2017. Springer Berlin - Heidelberg, 2017.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Vogel, C & Pock, T 2017, A Primal Dual Network for Low-Level Vision Problems. in German Conference on Pattern Recognition, 2017. Springer Berlin - Heidelberg.
Vogel C, Pock T. A Primal Dual Network for Low-Level Vision Problems. In German Conference on Pattern Recognition, 2017. Springer Berlin - Heidelberg. 2017.
Vogel, Christoph ; Pock, Thomas. / A Primal Dual Network for Low-Level Vision Problems. German Conference on Pattern Recognition, 2017. Springer Berlin - Heidelberg, 2017.
@inproceedings{87313210900541bd89a67ba197961e31,
title = "A Primal Dual Network for Low-Level Vision Problems",
abstract = "In the past, classic energy optimization techniques were the driving force in many innovations and are a building block for almost any problem in computer vision. Efficient algorithms are mandatory to achieve real-time processing, needed in many applications like autonomous driving. However, energy models - even if designed by human experts - might never be able to fully capture the complexity of natural scenes and images. Similar to optimization techniques, Deep Learning has changed the landscape of computer vision in recent years and has helped to push the performance of many models to never experienced heights.Our idea of a primal-dual network is to combine the structure of regular energy optimization techniques, in particular of first order methods, with the flexibility of Deep Learning to adapt to the statistics of the input data.",
author = "Christoph Vogel and Thomas Pock",
year = "2017",
month = "9",
day = "1",
language = "deutsch",
booktitle = "German Conference on Pattern Recognition, 2017",
publisher = "Springer Berlin - Heidelberg",

}

TY - GEN

T1 - A Primal Dual Network for Low-Level Vision Problems

AU - Vogel,Christoph

AU - Pock,Thomas

PY - 2017/9/1

Y1 - 2017/9/1

N2 - In the past, classic energy optimization techniques were the driving force in many innovations and are a building block for almost any problem in computer vision. Efficient algorithms are mandatory to achieve real-time processing, needed in many applications like autonomous driving. However, energy models - even if designed by human experts - might never be able to fully capture the complexity of natural scenes and images. Similar to optimization techniques, Deep Learning has changed the landscape of computer vision in recent years and has helped to push the performance of many models to never experienced heights.Our idea of a primal-dual network is to combine the structure of regular energy optimization techniques, in particular of first order methods, with the flexibility of Deep Learning to adapt to the statistics of the input data.

AB - In the past, classic energy optimization techniques were the driving force in many innovations and are a building block for almost any problem in computer vision. Efficient algorithms are mandatory to achieve real-time processing, needed in many applications like autonomous driving. However, energy models - even if designed by human experts - might never be able to fully capture the complexity of natural scenes and images. Similar to optimization techniques, Deep Learning has changed the landscape of computer vision in recent years and has helped to push the performance of many models to never experienced heights.Our idea of a primal-dual network is to combine the structure of regular energy optimization techniques, in particular of first order methods, with the flexibility of Deep Learning to adapt to the statistics of the input data.

M3 - Beitrag in einem Konferenzband

BT - German Conference on Pattern Recognition, 2017

PB - Springer Berlin - Heidelberg

ER -