Library
home (Home) TU Delft Repository login Login

Exploiting Learned Symmetries in Group Equivariant Convolutions

Conference Paper (2021)
Author(s)

Attila Lengyel (TU Delft - Electrical Engineering, Mathematics and Computer Science)

Jan van Gemert (TU Delft - Electrical Engineering, Mathematics and Computer Science)

Research Group
Pattern Recognition and Bioinformatics
DOI related publication
https://doi.org/10.1109/ICIP42928.2021.9506362 Final published version
More Info
expand_more
Publication Year
2021
Language
English
Related content
Research Group
Pattern Recognition and Bioinformatics
Article number
9506362
Pages (from-to)
759-763
ISBN (print)
978-1-6654-3102-6
ISBN (electronic)
978-1-6654-4115-5
Event
2021 IEEE International Conference on Image Processing (ICIP) (2021-09-19 - 2021-09-22), Virtual at Anchorage, United States
Downloads counter
120

Abstract

Group Equivariant Convolutions (GConvs) enable convolutional neural networks to be equivariant to various transformation groups, but at an additional parameter and compute cost. We investigate the filter parameters learned by GConvs and find certain conditions under which they become highly redundant. We show that GConvs can be efficiently decomposed into depthwise separable convolutions while preserving equivariance properties and demonstrate improved performance and data efficiency on two datasets. All code is publicly available at github.com/Attila94/SepGrouPy.