Off-Line Tamil Handwritten Character Recognition Based on Convolutional Neural Network with VGG16 and VGG19 Model

摘要

Automated character recognition is an evolving field of research that recognizes the character with the help of imaging technology. Deep learning pipeline have become a very popular method for various handwritten character recognition tasks. Over time, these network designs have mature to comprise many layers and are skilled of producing strong networks for character recognition tasks. Numerous Convolutional Neural Network (CNN) design have been planned such as LeNet, AlexNet, and GoogleNet depends on accuracy for various language character recognition. In this paper we in detail explore the usefulness of the VGG16 and VGG19 architecture on 25 class subset of the HP lab Offline Tamil isolated character dataset. We examine the benefits of transfer learning by using VGG network weights trained on the ImageNet dataset, which can provide better performance in the Tamil handwritten character recognition Task.