Deep Learning-Based Recognition of Handwritten English Characters

This paper presents a Convolutional Neural Network (CNN) model for handwritten character recognition. The architecture comprises three convolutional layers with ReLU activation functions with increased filter sizes like 32, 64 and 256 with max pooling, designed to extract relevant features. The model transitions to dense layers and culminates in a SoftMax output layer for classifying 26-character classes. To enhance performance and prevent overfitting, we implement dropout regularization, data augmentation techniques, early stopping, and learning rate reduction. The results demonstrate high accuracy and generalization, underscoring the efficacy of CNNs in image recognition tasks. This work contributes to the advancement of automatic handwritten character recognition, with potential applications in document digitization and educational technology.