A Novel Deep Learning Framework for Kannada Handwritten Character Recognition

Handwritten character recognition is a crucial process that involves converting handwritten text on different surfaces such as paper and postcards into digital formats, making it distinguishable from scanned images. Remarkable advancements have been made in this field, particularly in India, where many languages with complex scripts like Kannada, English, and Marathi are used. Kannada is one of the widely used languages in the southern part of India, and recognizing its handwritten characters is a challenging one task due to its complicated script. The article explains how CNN, a type of dense net, is used to recognize Kannada’s handwritten characters. This technology can help protect many historical documents from damage and destruction, a significant achievement. The article also highlights that the versatility of this machine learning model extends beyond Kannada and can be used in various applications and technology development. The proposed research has used CNN and transfer learning due to its efficient layers, and the data set from Kaggle consisted of 16 classes of 25 images each for experimentation. The results showed a remarkable 97.50 percent accuracy in 50 epochs, demonstrating the potential of the CNN network in character recognition.