<![CDATA[Handwritten numeral recognition is an important challenge in image processing, with wide applications in areas such as document processing and data automation. This research aims to optimize the performance of Convolutional Neural Network (CNN) model in classifying handwritten numerals on MNIST dataset. In this research, experiments were conducted with variations in the number of CNN layers to evaluate their effect on model accuracy. The results show that the model with 4 convolutional layers achieves the highest accuracy of 92.41%, which signifies a significant improvement in the model's ability to extract important features from the image compared to the model with fewer layers. This research also applied the best model to a website that allows users to recognize handwritten numerals in real-time. This provides practical benefits in the development of automatic character recognition systems and shows how this technology can be applied directly in everyday life.]]>
Copyrights © 2025
