Kannada Numerals Recognition and Translate into English Digits

Over the past few years, the discovery of handwritten numbers has become a research question in Optical Character Recognition (OCR).In this article we propose a CRNN model, which is a hybrid of convolutional neural networks (CNN) and recurrent neural networks (RNN).The GRU is used to replace the fully connected layer of the CNN in this architecture and improves the recognition accuracy.First, the CNN is excellent at extracting the most relevant features of the original image through a combination of convolutional and pooling layers.Secondly, GRU is a fast recognition algorithm, which forms a chain relation between the features in the hidden layer.The experimental studies carried out on the Kannada MNIST handwritten digit data demonstrate that the training accuracy of 98.12 and the validation accuracy of 98.95 are reached and the precision, recall, and F1 scores ratios are improved.Abstract--Handwritten character recognition is a significant field of Computer Vision that can help correct the human-machines interface.Machine learning and deep-learning methodologies have already reached acceptable levels of results, still, it can be improved.Offline Kannada handwritten character recognition is another field that many researchers are interested in; but the result of this has been only fairly satisfactory.