Hindi Handwritten Character Recognition Using Deep Neural Network

Handwritten character affirmation is a huge issue of account examination and affirmation. Composing has seen diverse such works that do this endeavour. A larger piece of this work exists for Latin, Chinese, and Arabic anyway unequivocally fewer works exist for Hindi substance. This hypothesis is an undertaking towards thinking about existing work and develop new methodologies to improve the exactness of separated interpreted Hindi character affirmation structures. A proposed incorporate extraction methodology, to be explicit frontal zone sub-examining (FS), which relies upon the level and vertical projection computation at each granularity level to find the division canters or feature canters. We further proposed a methodology through which the estimation of level and vertical projection at each granularity level ends up being brisk and capable by using vertical and even central pictures. If the model picture is 90 by 90 estimated by FS procedure at granularity level 3, 62100 extension (+) errands are expected to find 85 division canters, while in our proposed strategy only 18000 increments (+) exercises are adequate to deal with comparative features.