FILTRA: Rethinking Steerable CNN by Filter Transform

Bo Li, Qili Wang, Gim Hee Lee
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:6515-6522, 2021.

Abstract

Steerable CNN imposes the prior knowledge of transformation invariance or equivariance in the network architecture to enhance the the network robustness on geometry transformation of data and reduce overfitting. It has been an intuitive and widely used technique to construct a steerable filter by augmenting a filter with its transformed copies in the past decades, which is named as filter transform in this paper. Recently, the problem of steerable CNN has been studied from aspect of group representation theory, which reveals the function space structure of a steerable kernel function. However, it is not yet clear on how this theory is related to the filter transform technique. In this paper, we show that kernel constructed by filter transform can also be interpreted in the group representation theory. This interpretation help complete the puzzle of steerable CNN theory and provides a novel and simple approach to implement steerable convolution operators. Experiments are executed on multiple datasets to verify the feasibility of the proposed approach.

Cite this Paper

BibTeX
@InProceedings{pmlr-v139-li21v, title = {FILTRA: Rethinking Steerable CNN by Filter Transform}, author = {Li, Bo and Wang, Qili and Lee, Gim Hee}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {6515--6522}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/li21v/li21v.pdf}, url = {https://proceedings.mlr.press/v139/li21v.html}, abstract = {Steerable CNN imposes the prior knowledge of transformation invariance or equivariance in the network architecture to enhance the the network robustness on geometry transformation of data and reduce overfitting. It has been an intuitive and widely used technique to construct a steerable filter by augmenting a filter with its transformed copies in the past decades, which is named as filter transform in this paper. Recently, the problem of steerable CNN has been studied from aspect of group representation theory, which reveals the function space structure of a steerable kernel function. However, it is not yet clear on how this theory is related to the filter transform technique. In this paper, we show that kernel constructed by filter transform can also be interpreted in the group representation theory. This interpretation help complete the puzzle of steerable CNN theory and provides a novel and simple approach to implement steerable convolution operators. Experiments are executed on multiple datasets to verify the feasibility of the proposed approach.} }
Endnote
%0 Conference Paper %T FILTRA: Rethinking Steerable CNN by Filter Transform %A Bo Li %A Qili Wang %A Gim Hee Lee %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-li21v %I PMLR %P 6515--6522 %U https://proceedings.mlr.press/v139/li21v.html %V 139 %X Steerable CNN imposes the prior knowledge of transformation invariance or equivariance in the network architecture to enhance the the network robustness on geometry transformation of data and reduce overfitting. It has been an intuitive and widely used technique to construct a steerable filter by augmenting a filter with its transformed copies in the past decades, which is named as filter transform in this paper. Recently, the problem of steerable CNN has been studied from aspect of group representation theory, which reveals the function space structure of a steerable kernel function. However, it is not yet clear on how this theory is related to the filter transform technique. In this paper, we show that kernel constructed by filter transform can also be interpreted in the group representation theory. This interpretation help complete the puzzle of steerable CNN theory and provides a novel and simple approach to implement steerable convolution operators. Experiments are executed on multiple datasets to verify the feasibility of the proposed approach.
APA
Li, B., Wang, Q. & Lee, G.H.. (2021). FILTRA: Rethinking Steerable CNN by Filter Transform. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:6515-6522 Available from https://proceedings.mlr.press/v139/li21v.html.

Related Material