Classification of Written Numbers using 2D Convolutional Neural Network

Authors

  • Matúš Ďurovič Slovak University of Technology Author
  • Matúš Vaňo Faculty of Electrical Engineering and Information Technology, Slovak University of Technology in Bratislava, Slovakia Author
  • Juraj Kačur Faculty of Electrical Engineering and Information Technology, Slovak University of Technology in Bratislava, Slovakia Author

Keywords:

neural networks, convolutional neural networks, machine learning, image recognition, handwritten digits, parameter tuning

Abstract

This article introduces a configurable system for building CNNs designed for grayscale image classification. The network architecture and training settings are defined via a JSON file, allowing full control over the layers and parameters without modifying the code. The system was tested on MNIST dataset using two CNN configurations to classify handwritten numbers. Both achieved satisfactory accuracy. The modular design makes it suitable for rapid prototyping and experimentation in computer vision applications.

References

Downloads

Published

22.05.2025

Issue

Redžúr 2025 Proceedings

Section

Articles

How to Cite

[1]
M. Ďurovič, M. Vaňo, and J. Kačur, “Classification of Written Numbers using 2D Convolutional Neural Network”, R, vol. 17, pp. 77–82, May 2025, Accessed: May 08, 2026. [Online]. Available: https://redzur.stuba.sk/conf/article/view/17