Hand Written Indian Numeral Character Recognition using Deep Learning approaches
Hand written Character recognition is evolving topic as the size and shape of the hand written characters could not be uniquely characterized.
The proposed work considers the numeral characters in the image form in varieties of orientation and shape for the digits. The features are extracted and subjected to NaïveBayes classifier, BayesNet classifier and the results associated with the metrics are worked out. Recently, the deployment of machine learning techniques is widespread and has proven improved performance.
In the proposed work we have deployed the deep learning network with one stage of convolution layer with 20 numbers of 5×5 filters, Rectifier Linear Unit, Max pooling layer, fully connected softmax layer. The MNIST dataset with 60000 number of training images and 10000 numbers of testing images are used to experiment the proposed network.
The results observed using the deep learning techniques are superior to the existing classifier such as NaïveBayes and BayesNet classifier and also the recent work proposed using deep learning techniques with 3×3 filers in the convolution layer.