An Investigation of Methods for Improving Spatial Invariance of Convolutional Neural Networks for Image Classification

David Noel; Sumitra Mukherjee

doi:10.1007/978-3-031-86623-4_3

An Investigation of Methods for Improving Spatial Invariance of Convolutional Neural Networks for Image Classification

David Noel
, Sumitra Mukherjee

Nova Southeastern University

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

1 Scopus citations

Abstract

Convolutional Neural Networks are widely used for image recognition tasks. Data augmentation is a technique used to improve spatial invariance and reduce overfitting in Convolutional Networks. This article provides an empirical comparison of two common data augmentation techniques: RandAugment (a stochastic technique that applies geometric transforms to images) and Conditional Generative Adversarial Networks (models that generate synthetic samples from the same distribution as the training set). Combinations of these data augmentation techniques are also investigated. Three models – a base model developed for this study, and finetuned pre-trained versions of ResNet50 and InceptionV3 – are evaluated on benchmark datasets. Results indicate that RandAugment is more effective.

Original language	English
Title of host publication	Artificial Intelligence and Applications - 26th International Conference, ICAI 2024, Held as Part of the World Congress in Computer Science, Computer Engineering and Applied Computing, CSCE 2024, Revised Selected Papers
Editors	Hamid R. Arabnia, Leonidas Deligiannidis, Soheyla Amirian, Farzan Shenavarmasouleh, Farid Ghareh Mohammadi, David de la Fuente
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	33-50
Number of pages	18
ISBN (Print)	9783031866227
DOIs	https://doi.org/10.1007/978-3-031-86623-4_3
State	Published - 2025
Event	26th International Conference on Artificial Intelligence and Applications, ICAI 2024, held as part of the World Congress in Computer Science, Computer Engineering and Applied Computing, CSCE 2024 - Las Vegas, United States Duration: Jul 22 2024 → Jul 25 2024

Publication series

Name	Communications in Computer and Information Science
Volume	2252 CCIS
ISSN (Print)	1865-0929
ISSN (Electronic)	1865-0937

Conference

Conference	26th International Conference on Artificial Intelligence and Applications, ICAI 2024, held as part of the World Congress in Computer Science, Computer Engineering and Applied Computing, CSCE 2024
Country/Territory	United States
City	Las Vegas
Period	7/22/24 → 7/25/24

Bibliographical note

Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

ASJC Scopus Subject Areas

General Computer Science
General Mathematics

Keywords

convolutional neural networks
deep learning
generative adversarial networks
randaugment
spatial invariance

Access to Document

10.1007/978-3-031-86623-4_3

Cite this

Noel, D., & Mukherjee, S. (2025). An Investigation of Methods for Improving Spatial Invariance of Convolutional Neural Networks for Image Classification. In H. R. Arabnia, L. Deligiannidis, S. Amirian, F. Shenavarmasouleh, F. Ghareh Mohammadi, & D. de la Fuente (Eds.), Artificial Intelligence and Applications - 26th International Conference, ICAI 2024, Held as Part of the World Congress in Computer Science, Computer Engineering and Applied Computing, CSCE 2024, Revised Selected Papers (pp. 33-50). (Communications in Computer and Information Science; Vol. 2252 CCIS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-86623-4_3

An Investigation of Methods for Improving Spatial Invariance of Convolutional Neural Networks for Image Classification. / Noel, David; Mukherjee, Sumitra.
Artificial Intelligence and Applications - 26th International Conference, ICAI 2024, Held as Part of the World Congress in Computer Science, Computer Engineering and Applied Computing, CSCE 2024, Revised Selected Papers. ed. / Hamid R. Arabnia; Leonidas Deligiannidis; Soheyla Amirian; Farzan Shenavarmasouleh; Farid Ghareh Mohammadi; David de la Fuente. Springer Science and Business Media Deutschland GmbH, 2025. p. 33-50 (Communications in Computer and Information Science; Vol. 2252 CCIS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Noel, D & Mukherjee, S 2025, An Investigation of Methods for Improving Spatial Invariance of Convolutional Neural Networks for Image Classification. in HR Arabnia, L Deligiannidis, S Amirian, F Shenavarmasouleh, F Ghareh Mohammadi & D de la Fuente (eds), Artificial Intelligence and Applications - 26th International Conference, ICAI 2024, Held as Part of the World Congress in Computer Science, Computer Engineering and Applied Computing, CSCE 2024, Revised Selected Papers. Communications in Computer and Information Science, vol. 2252 CCIS, Springer Science and Business Media Deutschland GmbH, pp. 33-50, 26th International Conference on Artificial Intelligence and Applications, ICAI 2024, held as part of the World Congress in Computer Science, Computer Engineering and Applied Computing, CSCE 2024, Las Vegas, United States, 7/22/24. https://doi.org/10.1007/978-3-031-86623-4_3

Noel D, Mukherjee S. An Investigation of Methods for Improving Spatial Invariance of Convolutional Neural Networks for Image Classification. In Arabnia HR, Deligiannidis L, Amirian S, Shenavarmasouleh F, Ghareh Mohammadi F, de la Fuente D, editors, Artificial Intelligence and Applications - 26th International Conference, ICAI 2024, Held as Part of the World Congress in Computer Science, Computer Engineering and Applied Computing, CSCE 2024, Revised Selected Papers. Springer Science and Business Media Deutschland GmbH. 2025. p. 33-50. (Communications in Computer and Information Science). doi: 10.1007/978-3-031-86623-4_3

Noel, David ; Mukherjee, Sumitra. / An Investigation of Methods for Improving Spatial Invariance of Convolutional Neural Networks for Image Classification. Artificial Intelligence and Applications - 26th International Conference, ICAI 2024, Held as Part of the World Congress in Computer Science, Computer Engineering and Applied Computing, CSCE 2024, Revised Selected Papers. editor / Hamid R. Arabnia ; Leonidas Deligiannidis ; Soheyla Amirian ; Farzan Shenavarmasouleh ; Farid Ghareh Mohammadi ; David de la Fuente. Springer Science and Business Media Deutschland GmbH, 2025. pp. 33-50 (Communications in Computer and Information Science).

@inproceedings{3e495379247f42f6b909688dcaa1f75d,

title = "An Investigation of Methods for Improving Spatial Invariance of Convolutional Neural Networks for Image Classification",

abstract = "Convolutional Neural Networks are widely used for image recognition tasks. Data augmentation is a technique used to improve spatial invariance and reduce overfitting in Convolutional Networks. This article provides an empirical comparison of two common data augmentation techniques: RandAugment (a stochastic technique that applies geometric transforms to images) and Conditional Generative Adversarial Networks (models that generate synthetic samples from the same distribution as the training set). Combinations of these data augmentation techniques are also investigated. Three models {\textendash} a base model developed for this study, and finetuned pre-trained versions of ResNet50 and InceptionV3 {\textendash} are evaluated on benchmark datasets. Results indicate that RandAugment is more effective.",

keywords = "convolutional neural networks, deep learning, generative adversarial networks, randaugment, spatial invariance",

author = "David Noel and Sumitra Mukherjee",

note = "Publisher Copyright: {\textcopyright} The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.; 26th International Conference on Artificial Intelligence and Applications, ICAI 2024, held as part of the World Congress in Computer Science, Computer Engineering and Applied Computing, CSCE 2024 ; Conference date: 22-07-2024 Through 25-07-2024",

year = "2025",

doi = "10.1007/978-3-031-86623-4\_3",

language = "English",

isbn = "9783031866227",

series = "Communications in Computer and Information Science",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "33--50",

editor = "Arabnia, \{Hamid R.\} and Leonidas Deligiannidis and Soheyla Amirian and Farzan Shenavarmasouleh and \{Ghareh Mohammadi\}, Farid and \{de la Fuente\}, David",

booktitle = "Artificial Intelligence and Applications - 26th International Conference, ICAI 2024, Held as Part of the World Congress in Computer Science, Computer Engineering and Applied Computing, CSCE 2024, Revised Selected Papers",

address = "Germany",

}

TY - GEN

T1 - An Investigation of Methods for Improving Spatial Invariance of Convolutional Neural Networks for Image Classification

AU - Noel, David

AU - Mukherjee, Sumitra

N1 - Publisher Copyright: © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

PY - 2025

Y1 - 2025

N2 - Convolutional Neural Networks are widely used for image recognition tasks. Data augmentation is a technique used to improve spatial invariance and reduce overfitting in Convolutional Networks. This article provides an empirical comparison of two common data augmentation techniques: RandAugment (a stochastic technique that applies geometric transforms to images) and Conditional Generative Adversarial Networks (models that generate synthetic samples from the same distribution as the training set). Combinations of these data augmentation techniques are also investigated. Three models – a base model developed for this study, and finetuned pre-trained versions of ResNet50 and InceptionV3 – are evaluated on benchmark datasets. Results indicate that RandAugment is more effective.

AB - Convolutional Neural Networks are widely used for image recognition tasks. Data augmentation is a technique used to improve spatial invariance and reduce overfitting in Convolutional Networks. This article provides an empirical comparison of two common data augmentation techniques: RandAugment (a stochastic technique that applies geometric transforms to images) and Conditional Generative Adversarial Networks (models that generate synthetic samples from the same distribution as the training set). Combinations of these data augmentation techniques are also investigated. Three models – a base model developed for this study, and finetuned pre-trained versions of ResNet50 and InceptionV3 – are evaluated on benchmark datasets. Results indicate that RandAugment is more effective.

KW - convolutional neural networks

KW - deep learning

KW - generative adversarial networks

KW - randaugment

KW - spatial invariance

UR - https://www.scopus.com/pages/publications/105005251138

UR - https://www.scopus.com/pages/publications/105005251138#tab=citedBy

U2 - 10.1007/978-3-031-86623-4_3

DO - 10.1007/978-3-031-86623-4_3

M3 - Conference contribution

AN - SCOPUS:105005251138

SN - 9783031866227

T3 - Communications in Computer and Information Science

SP - 33

EP - 50

BT - Artificial Intelligence and Applications - 26th International Conference, ICAI 2024, Held as Part of the World Congress in Computer Science, Computer Engineering and Applied Computing, CSCE 2024, Revised Selected Papers

A2 - Arabnia, Hamid R.

A2 - Deligiannidis, Leonidas

A2 - Amirian, Soheyla

A2 - Shenavarmasouleh, Farzan

A2 - Ghareh Mohammadi, Farid

A2 - de la Fuente, David

PB - Springer Science and Business Media Deutschland GmbH

T2 - 26th International Conference on Artificial Intelligence and Applications, ICAI 2024, held as part of the World Congress in Computer Science, Computer Engineering and Applied Computing, CSCE 2024

Y2 - 22 July 2024 through 25 July 2024

ER -

An Investigation of Methods for Improving Spatial Invariance of Convolutional Neural Networks for Image Classification

Abstract

Publication series

Conference

Bibliographical note

ASJC Scopus Subject Areas

Keywords

Access to Document

Other files and links

Fingerprint

Cite this