Enhancing Tamil Handwritten Character Recognition Using Multimodel Deep Learning

Abstract

Based of the intricate script and lengthy history of the language, handwritten character identification in Tamil presents particular difficulties. In order to improve the accuracy of Tamil handwritten character identification, this research suggests a unique method that makes use of multimodel deep learning techniques. In response to the shortcomings of the current datasets, we provide the unconstrained Tamil Handwritten Character Database (uTHCD) dataset. RESNET50, VGG16, and LeNet50 architectures are used in our methods to extract features and classify them. We establish the effectiveness of our method by conducting a large-scale experiment and attaining notable gains in recognition accuracy. Our results highlight the value of multimodel learning for handwritten Tamil documents preservation and digitization, enabling their smooth incorporation into digital settings.