Deep Learning for 2D and 3D Rotatable Data: An Overview of Methods
Abstract
Convolutional networks are successful due to their equivariance/invariance under translations. However, rotatable data such as images, volumes, shapes, or point clouds require processing with equivariance/invariance under rotations in cases where the rotational orientation 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.
Cite
@article{arxiv.1910.14594,
title = {Deep Learning for 2D and 3D Rotatable Data: An Overview of Methods},
author = {Luca Della Libera and Vladimir Golkov and Yue Zhu and Arman Mielke and Daniel Cremers},
journal= {arXiv preprint arXiv:1910.14594},
year = {2021}
}
Comments
Improved Definition 1, improved and merged Sections 3.3-3.4, minor additional changes