IEICE TRANSACTIONS on Information
Rotation-Invariant Convolution Networks with Hexagon-Based Kernels
Summary :
We introduce the Hexagonal Convolutional Neural Network (HCNN), a modified version of CNN that is robust against rotation. HCNN utilizes a hexagonal kernel and a multi-block structure that enjoys more degrees of rotation information sharing than standard convolution layers. Our structure is easy to use and does not affect the original tissue structure of the network. We achieve the complete rotational invariance on the recognition task of simple pattern images and demonstrate better performance on the recognition task of the rotated MNIST images, synthetic biomarker images and microscopic cell images than past methods, where the robustness to rotation matters.
- Publication
- IEICE TRANSACTIONS on Information Vol.E107-D No.2 pp.220-228
- Publication Date
- 2024/02/01
- Publicized
- 2023/11/15
- Online ISSN
- 1745-1361
- DOI
- 10.1587/transinf.2023EDP7023
- Type of Manuscript
- PAPER
- Category
- Biocybernetics, Neurocomputing
Authors
Yiping TANG
Kyushu University
Kohei HATANO
Kyushu University,Riken AIP
Eiji TAKIMOTO
Kyushu University
Keyword
Latest Issue
Copyrights notice of machine-translated contents
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Cite this
Copy
Yiping TANG, Kohei HATANO, Eiji TAKIMOTO, "Rotation-Invariant Convolution Networks with Hexagon-Based Kernels" in IEICE TRANSACTIONS on Information,
vol. E107-D, no. 2, pp. 220-228, February 2024, doi: 10.1587/transinf.2023EDP7023.
Abstract: We introduce the Hexagonal Convolutional Neural Network (HCNN), a modified version of CNN that is robust against rotation. HCNN utilizes a hexagonal kernel and a multi-block structure that enjoys more degrees of rotation information sharing than standard convolution layers. Our structure is easy to use and does not affect the original tissue structure of the network. We achieve the complete rotational invariance on the recognition task of simple pattern images and demonstrate better performance on the recognition task of the rotated MNIST images, synthetic biomarker images and microscopic cell images than past methods, where the robustness to rotation matters.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2023EDP7023/_p
Copy
@ARTICLE{e107-d_2_220,
author={Yiping TANG, Kohei HATANO, Eiji TAKIMOTO, },
journal={IEICE TRANSACTIONS on Information},
title={Rotation-Invariant Convolution Networks with Hexagon-Based Kernels},
year={2024},
volume={E107-D},
number={2},
pages={220-228},
abstract={We introduce the Hexagonal Convolutional Neural Network (HCNN), a modified version of CNN that is robust against rotation. HCNN utilizes a hexagonal kernel and a multi-block structure that enjoys more degrees of rotation information sharing than standard convolution layers. Our structure is easy to use and does not affect the original tissue structure of the network. We achieve the complete rotational invariance on the recognition task of simple pattern images and demonstrate better performance on the recognition task of the rotated MNIST images, synthetic biomarker images and microscopic cell images than past methods, where the robustness to rotation matters.},
keywords={},
doi={10.1587/transinf.2023EDP7023},
ISSN={1745-1361},
month={February},}
Copy
TY - JOUR
TI - Rotation-Invariant Convolution Networks with Hexagon-Based Kernels
T2 - IEICE TRANSACTIONS on Information
SP - 220
EP - 228
AU - Yiping TANG
AU - Kohei HATANO
AU - Eiji TAKIMOTO
PY - 2024
DO - 10.1587/transinf.2023EDP7023
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E107-D
IS - 2
JA - IEICE TRANSACTIONS on Information
Y1 - February 2024
AB - We introduce the Hexagonal Convolutional Neural Network (HCNN), a modified version of CNN that is robust against rotation. HCNN utilizes a hexagonal kernel and a multi-block structure that enjoys more degrees of rotation information sharing than standard convolution layers. Our structure is easy to use and does not affect the original tissue structure of the network. We achieve the complete rotational invariance on the recognition task of simple pattern images and demonstrate better performance on the recognition task of the rotated MNIST images, synthetic biomarker images and microscopic cell images than past methods, where the robustness to rotation matters.
ER -
English
