IMPLEMENTATION OF CONVOLUTIONAL NEURAL NETWORK FOR SUNDANESE SCRIPT HANDWRITING RECOGNITION WITH DATA AUGMENTATION

Sundanese script is one of the cultural heritages that need to be preserved. However, Sundanese script has complexity and uniqueness in its writing, making it difficult to recognize. The recognition can be done automatically using deep learning. One of the problems is that the recognition has a small amount of data and is less varied. In this study, the proposed solution is to use data augmentation. This study focused on how the use of data augmentation can help to improve accuracy in performing the recognition of handwriting image pattern using Convolutional Neural Network (CNN) method. Data augmentation is the process of artificially increasing the amount of data by generating new data points from existing data. The augmentation includes adding small changes to the data or using machine learning models to generate new data points in the latent space of the original data in order to strengthen the data set. Data augmentation applied in this research is flipping, rotation, and translation techniques. Based on the results, it can be concluded that the use of data augmentation to increase the number and variety of data samples has a significant effect on the accuracy value of 0.1707 or about 17.07%. The best accuracy obtained is 0.8 or 80% using a baseline model with data augmentation. These findings yielded good results because the system is able to perform image classification quite well.