Tamil and English Handwritten Character Segmentation and Recognition Using Deep Learning

Handwritten character recognition is a critical component in various applications, ranging from text transcription to document analysis. This study focuses on the development of an innovative system that addresses the segmentation and recognition of handwritten characters in both Tamil and English scripts, leveraging the capabilities of scikit-learn, metrics, TensorFlow, and Keras within the realm of deep learning technologies. The research commences by employing advanced image processing techniques for character segmentation, facilitating the isolation of individual characters from the input text. Subsequently, a deep learning model is trained using the powerful combination of TensorFlow and Keras to recognize and classify these segmented characters. Throughout the experimental phase, scikit-learn is utilized for model evaluation, and various metrics are employed to assess its performance. The results highlight the system’s efficacy in character recognition for both Tamil and English scripts, positioning it as a valuable tool for multilingual character recognition, postal card digitization, and accessibility enhancement. Specifically, Accuracy for Tamil is reported as 65%, and for English, it is 85% in the output. The number of samples used for Tamil is 53,500, while for English, the number is 8,500. This project not only contributes to the broader field of optical character recognition but also emphasizes the potential of utilizing scikit-learn, metrics, TensorFlow, and Keras in handwritten script recognition within the context of the Post office.