Convolutional neural network (CNN) based month's name recognition in Gurumukhi script for Punjab state of India

In the context of natural language recognition, the development of an automated system for data analysis and interpretation has tremendous demand. Developing such kinds of systems for a country like India, however, has been found to be a laborious task when compared to other countries due to India's use of multiple scripts and languages. Gurumukhi is the regional language of Punjab, and the development of an automated text recognition system in the Gurumukhi language is found to be very critical because of the character's intricate structure. Influenced by this problem, the present work has been conducted to design an error-free classification model of Gurumukhi text. A convolutional neural network (CNN) based classification model was designed to perform holistic word recognition of Gurumukhi months. For this work, a dataset of 24,000 word images of 24 different Gurumukhi months has also been prepared from 500 distinct writers. The highest accuracy obtained using the proposed model on validation data is 99.7%.