Abstract
Handwritten text recognition is a challenging task with significant implications for cultural preservation and technological advancement. In this research paper, we propose a unique approach for Hindko Handwritten text recognition. The primary objectives of the study were to develop an accurate and efficient system capable of recognizing Hindko handwritten text and to compare the performance of deep learning models, Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) and HTR-Attention model. To achieve these objectives, we collected a substantial dataset of handwritten Hindko text, comprising diverse writing styles and variations. Processing techniques were applied to enhance the data quality and extract meaningful features for training the deep neural networks. Through rigorous experimentation and evaluation, the research findings demonstrated promising results. The HTR-Attention model achieved and impressive accuracy of 88.97%. while CNN and RNN models attained an accuracy of 88.36% and 80.45% respectively. The potential applications of this research encompass various fields, including language recognition systems, optical character recognition systems, and language translation technologies. Furthermore, the developed model can be integrated into digital learning platforms to facilitate Hindko language learners and preserve the cultural heritage associated with this regional Language.