A Novel ConvNet Architecture for Recognizing Offline Handwritten Gujarati Conjuncts

Convolutional neural networks have emerged as a crucial tool in computer vision applications. In the present study, we have proposed a novel convolutional neural network model for the recognition of handwritten conjuncts of the Gujarati script. The model is developed from scratch and trained using the participating preceding and succeeding components of the conjuncts. Prior to training, meticulous segmentation of conjuncts is performed to retrieve these components. The scope of this study includes 767 distinct conjuncts, providing a robust foundation for evaluating the model's performance across a varied set of linguistic contexts. The recognition accuracy attained for the preceding components of the conjuncts is 95.66%, while the succeeding components exhibit a recognition accuracy of 95.54%. The results of this study are significant as they demonstrate the effectiveness of a convolutional neural network model in recognizing complex handwritten conjuncts in an under-resourced language like Gujarati. It is an attempt to set a standard for recognizing offline handwritten Gujarati conjuncts by employing cutting-edge deep learning techniques.