<![CDATA[Regional languages and scripts, including the Melayu Ogan script, face the threat of extinction due to declining usage and limited digital documentation in the modern era. While current Indonesian script research primarily focuses on popular scripts, research addressing the Ogan Malay script remains severely limited. To address this gap, this study provides one of the earliest implementations of the Convolutional Neural Network (CNN) VGG-19 architecture specifically designed for Ogan Malay script classification. This research utilizes a primary dataset provided by the Language Center of South Sumatra Province, consisting of 185 distinct character classes, with each class initially containing one original image. The VGG-19 architecture is applied and supported by data augmentation techniques to enrich spatial variability, followed by evaluation using k-fold cross-validation. Evaluation results demonstrate excellent classification performance. The model achieved maximum convergence without any indications of overfitting at an optimal configuration of 30 epochs with a learning rate of 0.0001. This configuration successfully resulted in an accuracy of 99.14%, a precision of 0.9870, a recall of 0.9914, and an F1-score of 0.9885. The success of this classification model provides a strong foundation for future real-time regional script recognition applications to support cultural preservation.]]>
Copyrights © 2026
