Integrating CNN and Random Forests for Effective Punjabi Handwriting Classification and Language Detection
Simran, Rishabh Sharma, Manish Nagpal · 2024
Convolutional neural networks (CNNs) and Random Forests (RF) form the basis of the hybrid model used to create the Punjabi handwriting classification system. The model is applied to a real-time dataset of 8000 images in various Punjabi language styles, including Gurumukhi Script and Shahmukhi script. For each class, 4000 images have been gathered to aid in classification. An RF and a CNN are employed in the proposed feature extraction and classification work. Convolutional layers extract local patterns and feature hierarchies from the text. RF can classify handwritten letters based on information extracted from images, such as gradients and pixel values. In 93.27% of cases, language categorization has been finished with detection precision. With detection accuracy, language categorization has been completed. An improvement in Punjabi language identification detection precision will be demonstrated by conducting the same experiment with increasing degrees of severity using different image datasets. The study aims to improve Punjabi language recognition detection precision by developing a hybrid categorization model and integrating Punjabi language resources into global digital platforms, thereby promoting cultural diversity and linguistic inclusivity.