Abstract

This paper gives the results of a study of the possibilities of convolutional neural networks to generalize knowledge concerning primitive geometrical image transformations when solving pattern-recognition problems of handwritten numerals. Experiments were directed to the study of how the recognition of patterns in arbitrary orientations is affected by broadening the training sample with rotated images. Results are presented for convolutional neural networks of two architectures, showing that, to ensure rotation-invariant recognition, it is necessary for all classes of images in the entire range of rotations to be present in the training sample.

© 2015 Optical Society of America

Neural network model for rotation invariant recognition of object shapes

Mausumi Pohit
Appl. Opt. 49(22) 4144-4151 (2010)

Learning, invariance, and generalization in high-order neural networks

C. Lee Giles and Tom Maxwell
Appl. Opt. 26(23) 4972-4978 (1987)

Improvement of the generalization capability for a pattern-recognition neural network that uses a Gaussian-synapse neuron model

Xin Lin, Masahiko Mori, and Masanobu Watanabe
Appl. Opt. 39(5) 770-775 (2000)

Scale, translation, and rotation invariant orthonormalized optical/optoelectronic neural networks

Edwin Ghahramani and L. Roger B. Patterson
Appl. Opt. 32(35) 7225-7232 (1993)

Processing of chromatic information in a deep convolutional neural network

Alban Flachot and Karl R. Gegenfurtner
J. Opt. Soc. Am. A 35(4) B334-B346 (2018)