Premium
Combination of Convolutional Neural Network Architecture and its Learning Method for Rotation‐Invariant Handwritten Digit Recognition
Author(s) -
Urazoe Kazuya,
Kuroki Nobutaka,
Hirose Tetsuya,
Numa Masahiro
Publication year - 2021
Publication title -
ieej transactions on electrical and electronic engineering
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.254
H-Index - 30
eISSN - 1931-4981
pISSN - 1931-4973
DOI - 10.1002/tee.23278
Subject(s) - convolutional neural network , invariant (physics) , rotation (mathematics) , computer science , deep learning , artificial intelligence , pattern recognition (psychology) , digit recognition , speech recognition , numerical digit , artificial neural network , arithmetic , mathematics , mathematical physics
This letter presents several combinations of a convolutional neural network (CNN) and its learning method for rotation‐invariant digit recognition. Rotation data augmentation is widely used for improving rotation invariance. Data augmentation commonly assigns the same label to all augmented images of the same source. However, this learning method causes some collisions between original and rotated digits. Thus, this letter presents three types of rotation‐invariance learning methods and applies them to five popular CNN architectures. Experimental results indicate that multi‐task learning on ResNet‐50 is the best combination. © 2020 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC.