Title: Image identification using convolutional neural networks

Authors: Hameem Ahsan; Sohana Jahan; Md. Anwarul Islam Bhuiyan

Addresses: Department of Mathematics, University of Dhaka, Dhaka – 1000, Bangladesh ' Department of Mathematics, University of Dhaka, Dhaka – 1000, Bangladesh ' Department of Mathematics, University of Dhaka, Dhaka – 1000, Bangladesh

Abstract: Image recognition is an element of computer vision that includes techniques for detecting, processing, and classifying images. Due to its frequent use in artificial intelligence, the demand for better image recognition methods is increasing exponentially. Key spatial characteristics in image data can be identified through convolutional neural networks (CNN), a subset of deep learning. In this article, CNN-based frameworks for the MNIST, CIFAR-10, and cats vs. dogs datasets have been developed. A supervised learning technique is used to train the proposed models. The proposed dataset-specific method of using CNN to solve categorisation issues has produced remarkable results. Numerical experiments suggest that the efficacy of the proposed models is comparable to that of state-of-the-art CNN frameworks.

Keywords: image identification; CNN; regularisation; dropout; max pooling; loss function.

DOI: 10.1504/IJSSS.2024.144732

International Journal of Society Systems Science, 2024 Vol.15 No.2, pp.115 - 146

Received: 04 Aug 2023
Accepted: 07 Oct 2024
Published online: 28 Feb 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article