Tuning Convolutional Neural Networks Hyperparameters for Offline Handwriting Recognition

Ahmed Remaida, Aniss Moumen, Younes El Bouzekri El Idrissi, Benyoussef Abdellaoui

2021

Abstract

Deep Learning Artificial Neural Networks has pushed forward researches in the field of image recognition, furthermore in handwriting recognition. In writing or writer identification, segmentation, or features extraction applications, many ANNs models are applied in the process. This paper presents a comparative study of one of the most widely used ANNs for offline handwriting recognition, known as Deep Convolutional Neural Networks. We describe the challenging benchmark Dataset entitled EMNIST introduced in 2017 as an extended version of the well-known MNIST to fill the gap of handwritten letters characters. The accuracies obtained in this work for the letters dataset compares favourably with many other approaches in the literature. The effect of the choice of hyperparameters related to our network architecture and capabilities are explored and detailed, like the number of layers, neurons, optimizers, learning rates and other parameters.

Paper Citation

in Harvard Style

Remaida A., Moumen A., El Bouzekri El Idrissi Y. and Abdellaoui B. (2021). Tuning Convolutional Neural Networks Hyperparameters for Offline Handwriting Recognition. In Proceedings of the 2nd International Conference on Big Data, Modelling and Machine Learning - Volume 1: BML, ISBN 978-989-758-559-3, pages 71-76. DOI: 10.5220/0010728600003101

in Bibtex Style

@conference{bml21,
author={Ahmed Remaida and Aniss Moumen and Younes El Bouzekri El Idrissi and Benyoussef Abdellaoui},
title={Tuning Convolutional Neural Networks Hyperparameters for Offline Handwriting Recognition},
booktitle={Proceedings of the 2nd International Conference on Big Data, Modelling and Machine Learning - Volume 1: BML,},
year={2021},
pages={71-76},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0010728600003101},
isbn={978-989-758-559-3},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 2nd International Conference on Big Data, Modelling and Machine Learning - Volume 1: BML,
TI - Tuning Convolutional Neural Networks Hyperparameters for Offline Handwriting Recognition
SN - 978-989-758-559-3
AU - Remaida A.
AU - Moumen A.
AU - El Bouzekri El Idrissi Y.
AU - Abdellaoui B.
PY - 2021
SP - 71
EP - 76
DO - 10.5220/0010728600003101