Transformation of handwritten Devnagari script into word editable form using CNN

In India, many historical documents are preserved where handwritten Devnagari script has been used. With the advancement of time, the handwritten documents is in deforming styles and shapes. The handwritten OCR software that can transform this task with different accuracy rate. In 2015, this work was done for the printed character and provided the accuracy rate for Devanagari numerals of 89% with 4.5% of confusion rate whereas the recognition rate for English set was 78.4% with 18% of confusion rate. Although many work has been done towards HOCR but accuracy rate is 94% to 97%. My research work will be towards the transformation of handwritten Devanagari script into editable word form by using Convolutional Neural Network. For choosing this area for my research is because transformation of handwritten ancient characters is important as many important historical documents like manuscripts into machine editable form which can be easily preserved and accessed. With the help of strong and powerful dataset and combination of different neural network techniques, the accuracy rate can be increased and we can be able to preserve more historical documents for our access.