Abstract
In this paper, we address the task of Optical Character Recognition (OCR) for Telugu script. Telugu is a Dravidian language, native to India. OCR for English is well established and there are many mobile apps available. For Telugu the complexity is much more higher because of number of output classes possible and the inter class variability. Further there aren’t any good OCR systems for Telugu. The main aim of this paper is to make an accruate end to end solution of Telugu. Inspired by the success of CNN in Object recognition, Character recognition, Segmentation etc; we employ CNNs in character recognition for Telugu. We also employ an MSER based solution for character segmentation like in English with minor improvements. We made server-client based mobile app of OCR for Telugu so that it can be used by people with poor vision.
