Achieving Rotational Invariance with Bessel-Convolutional Neural Networks

Valentin Delchevalerie; Adrien Bibal; Benoît Frénay; Alexandre Mayer

Achieving Rotational Invariance with Bessel-Convolutional Neural Networks

Valentin Delchevalerie, Adrien Bibal, Benoît Frénay, Alexandre Mayer

Research output: Contribution in Book/Catalog/Report/Conference proceeding › Conference contribution

41 Downloads (Pure)

Abstract

For many applications in image analysis, learning models that are invariant to translations and rotations is paramount. This is the case, for example, in medical imaging where the objects of interest can appear at arbitrary positions, with arbitrary orientations. As of today, Convolutional Neural Networks (CNN) are one of the most powerful tools for image analysis. They achieve, thanks to convolutions, an invariance with respect to translations. In this work, we present a new type of convolutional layer that takes advantage of Bessel functions, well known in physics, to build Bessel-CNNs (B-CNNs) that are invariant to all the continuous set of possible rotation angles by design.

Original language	English
Title of host publication	35th Conference on Neural Information Processing Systems
Subtitle of host publication	NeurIPS 2021
Number of pages	12
Volume	34
ISBN (Electronic)	9781713845393
Publication status	Published - 2021
Event	35th Conference on Neural Information Processing Systems - Duration: 6 Dec 2021 → 14 Dec 2021 Conference number: 35

Conference

Conference	35th Conference on Neural Information Processing Systems
Abbreviated title	NeurIPS 2021
Period	6/12/21 → 14/12/21

Keywords

machine learning
computer vision
neural networks
image classification
rotational invariance
Bessel-CNN

Access to Document

NeurIPS_Bessel_CNN

https://proceedings.neurips.cc/paper/2021/hash/f18224a1adfb7b3dbff668c9b655a35a-Abstract.htmlLicence: Unspecified

CÉCI – Consortium of high performance computing centers
CHAMPAGNE, B. (PI), Lazzaroni, R. (PI), Geuzaine , C. (CoI), Chatelain, P. (CoI) & Knaepen, B. (CoI)
1/01/18 → 31/12/22
Project: Research

High Performance Computing Technology Platform
Champagne, B. (Manager)
Technological Platform High Performance Computing
Facility/equipment: Technological Platform

Cite this

@inproceedings{671d42c6f9b447ce8012024e9b0ab4ac,

title = "Achieving Rotational Invariance with Bessel-Convolutional Neural Networks",

abstract = "For many applications in image analysis, learning models that are invariant to translations and rotations is paramount. This is the case, for example, in medical imaging where the objects of interest can appear at arbitrary positions, with arbitrary orientations. As of today, Convolutional Neural Networks (CNN) are one of the most powerful tools for image analysis. They achieve, thanks to convolutions, an invariance with respect to translations. In this work, we present a new type of convolutional layer that takes advantage of Bessel functions, well known in physics, to build Bessel-CNNs (B-CNNs) that are invariant to all the continuous set of possible rotation angles by design.",

keywords = "machine learning, computer vision, neural networks, image classification, rotational invariance, Bessel-CNN",

author = "Valentin Delchevalerie and Adrien Bibal and Beno{\^i}t Fr{\'e}nay and Alexandre Mayer",

year = "2021",

language = "English",

volume = "34",

booktitle = "35th Conference on Neural Information Processing Systems",

note = "35th Conference on Neural Information Processing Systems ; Conference date: 06-12-2021 Through 14-12-2021",

}

Delchevalerie, V , Bibal, A , Frénay, B & Mayer, A 2021, Achieving Rotational Invariance with Bessel-Convolutional Neural Networks. in 35th Conference on Neural Information Processing Systems: NeurIPS 2021. vol. 34, 35th Conference on Neural Information Processing Systems, 6/12/21. <https://proceedings.neurips.cc/paper/2021/hash/f18224a1adfb7b3dbff668c9b655a35a-Abstract.html>

TY - GEN

T1 - Achieving Rotational Invariance with Bessel-Convolutional Neural Networks

AU - Delchevalerie, Valentin

AU - Bibal, Adrien

AU - Frénay, Benoît

AU - Mayer, Alexandre

N1 - Conference code: 35

PY - 2021

Y1 - 2021

N2 - For many applications in image analysis, learning models that are invariant to translations and rotations is paramount. This is the case, for example, in medical imaging where the objects of interest can appear at arbitrary positions, with arbitrary orientations. As of today, Convolutional Neural Networks (CNN) are one of the most powerful tools for image analysis. They achieve, thanks to convolutions, an invariance with respect to translations. In this work, we present a new type of convolutional layer that takes advantage of Bessel functions, well known in physics, to build Bessel-CNNs (B-CNNs) that are invariant to all the continuous set of possible rotation angles by design.

AB - For many applications in image analysis, learning models that are invariant to translations and rotations is paramount. This is the case, for example, in medical imaging where the objects of interest can appear at arbitrary positions, with arbitrary orientations. As of today, Convolutional Neural Networks (CNN) are one of the most powerful tools for image analysis. They achieve, thanks to convolutions, an invariance with respect to translations. In this work, we present a new type of convolutional layer that takes advantage of Bessel functions, well known in physics, to build Bessel-CNNs (B-CNNs) that are invariant to all the continuous set of possible rotation angles by design.

KW - machine learning

KW - computer vision

KW - neural networks

KW - image classification

KW - rotational invariance

KW - Bessel-CNN

M3 - Conference contribution

VL - 34

BT - 35th Conference on Neural Information Processing Systems

T2 - 35th Conference on Neural Information Processing Systems

Y2 - 6 December 2021 through 14 December 2021

ER -

Achieving Rotational Invariance with Bessel-Convolutional Neural Networks

Abstract

Conference

Keywords

Access to Document

Fingerprint

Projects

CÉCI – Consortium of high performance computing centers

Equipment

High Performance Computing Technology Platform

Cite this