Optical Recognition of the Philippines' Ancient Text: A Deep Learning Approach

Baybayin, an ancient writing system of the Philippines, has been largely forgotten due to the introduction of Latin-based scripts during colonization. However, with the increasing interest in cultural preservation, there is a need for a robust Optical Character Recognition (OCR) system that can accurately recognize and digitize Baybayin text. This study explores the development of a Baybayin OCR system using Convolutional Neural Networks (CNNs) to improve character recognition accuracy. The model is trained on a diverse dataset consisting of handwritten, printed, and synthetic Baybayin characters. Data preprocessing techniques such as image normalization, noise reduction, and data augmentation are applied to enhance model generalization. The CNN model is evaluated using accuracy, precision, recall, and F1-score, achieving an accuracy of 96.2%, significantly outperforming traditional OCR methods such as Tesseract. The trained model is further optimized using TensorFlow Lite (TFLite) for mobile deployment, enabling real-time character recognition via smartphone cameras. The findings of this study demonstrate the feasibility of deep learning-based OCR systems in recognizing ancient scripts, contributing to both cultural preservation and digital accessibility.