International Research Journal of Education and Technology

Palm leaf manuscripts are very significant because they provide a wealth of information. As a result, easy access to historical manuscripts must be made available in order to disseminate this knowledge around the world and encourage further research into ancient literature. In this study, an optical character recognition (OCR) system based on convolutional neural networks (CNN) is used to precisely digitize and identify Tamil palm leaf manuscript characters. In this article, the convolutional neural network's classifier, pooling layer, activation layer, fully connected layer, and other layers are used. The character set database was created using the scanned images of palm-leaf manuscripts. The database is split into 67 distinct classes, with about 100 samples in each class. The palm leaf manuscripts' OCR recognition and related issues are illustrated. The CNN model was used to construct a functioning example of the character recognition technique for Tamil palm-leaf text. It was discovered that the CNN model had a higher rate of recognition. Because so many features were extracted for each layer of CNN, the prediction rate and accuracy are very high.

      Keywords - Palm leaf manuscript,Digitization,Convolutional NueralNetworks,Classification