One of the reasons for the success of convolutional networks is their
equivariance/invariance under translations. However, rotatable data such as
molecules, living cells, everyday objects, or galaxies require processing with
equivariance/invariance under rotations in cases where the rotation of the
coordinate system does not affect the meaning of the data (e.g. object
classification). On the other hand, estimation/processing of rotations is
necessary in cases where rotations are important (e.g. motion estimation).
There has been recent progress in methods and theory in all these regards. Here
we provide an overview of existing methods, both for 2D and 3D rotations (and
translations), and identify commonalities and links between them, in the hope
that our insights will be useful for choosing and perfecting the methods.
Description
[1910.14594] Deep Learning for 2D and 3D Rotatable Data: An Overview of Methods
%0 Generic
%1 dellalibera2019learning
%A Della Libera, Luca
%A Golkov, Vladimir
%A Zhu, Yue
%A Mielke, Arman
%A Cremers, Daniel
%D 2019
%K 2019 deep-learning research review rotation
%T Deep Learning for 2D and 3D Rotatable Data: An Overview of Methods
%U http://arxiv.org/abs/1910.14594
%X One of the reasons for the success of convolutional networks is their
equivariance/invariance under translations. However, rotatable data such as
molecules, living cells, everyday objects, or galaxies require processing with
equivariance/invariance under rotations in cases where the rotation of the
coordinate system does not affect the meaning of the data (e.g. object
classification). On the other hand, estimation/processing of rotations is
necessary in cases where rotations are important (e.g. motion estimation).
There has been recent progress in methods and theory in all these regards. Here
we provide an overview of existing methods, both for 2D and 3D rotations (and
translations), and identify commonalities and links between them, in the hope
that our insights will be useful for choosing and perfecting the methods.
@misc{dellalibera2019learning,
abstract = {One of the reasons for the success of convolutional networks is their
equivariance/invariance under translations. However, rotatable data such as
molecules, living cells, everyday objects, or galaxies require processing with
equivariance/invariance under rotations in cases where the rotation of the
coordinate system does not affect the meaning of the data (e.g. object
classification). On the other hand, estimation/processing of rotations is
necessary in cases where rotations are important (e.g. motion estimation).
There has been recent progress in methods and theory in all these regards. Here
we provide an overview of existing methods, both for 2D and 3D rotations (and
translations), and identify commonalities and links between them, in the hope
that our insights will be useful for choosing and perfecting the methods.},
added-at = {2020-07-05T16:13:56.000+0200},
author = {Della Libera, Luca and Golkov, Vladimir and Zhu, Yue and Mielke, Arman and Cremers, Daniel},
biburl = {https://www.bibsonomy.org/bibtex/2f48533c6c09978b652b6699b688cade5/analyst},
description = {[1910.14594] Deep Learning for 2D and 3D Rotatable Data: An Overview of Methods},
interhash = {6b62bf35203c87298d70fd1941c6d06d},
intrahash = {f48533c6c09978b652b6699b688cade5},
keywords = {2019 deep-learning research review rotation},
note = {cite arxiv:1910.14594},
timestamp = {2020-07-05T16:13:56.000+0200},
title = {Deep Learning for 2D and 3D Rotatable Data: An Overview of Methods},
url = {http://arxiv.org/abs/1910.14594},
year = 2019
}