Persian Handwritten Character Recognition Using Convolutional Neural Network

While there has been some fascinating researches in Persian handwritten character recognition, there is still high demand for approaches with more accurate results. In order to achieve this goal, we use a deep learning architecture, in general, and Convolutional Neural Networks (CNN) in particular, for automatic feature extraction which is different from conventional hand-crafted feature extraction. Recently CNNs have solved many problems in sophisticated computer vision tasks. In the presented model, some optimization methods are used in order to enhance the performance of the CNN. The proposed model was trained and tested using the HODA dataset resulting an accuracy of 97.7%.