Isolated Telugu Language Character Recognition Using Deep Learning

The digital era digitizes every document. Handwritten documents are complex to digitize and analyze the content. Sophisticate character recognition systems are required to study the conceptual information in the documents. Telugu is one of the oldest languages, where most ancient records prevailed handwritten. A Telugu character recognition system is the need of the hour, which reveals most historic information to the modern community. Deep learning based architecture is proposed for the recognition of Telugu characters. The approach uses a vanilla model of convolutional neural network for the recognition of handwritten Telugu characters, by employing adaptive feature extraction. The proposed model has been executed on the large data sets collected from the physical libraries of Telugu Academy and best results are produced in both training and testing. The proposed model achieved 98% accuracy approximately which is a high figure compared to existing approaches for Telugu character recognition..