Recognition of Baybayin Symbols (Ancient Pre-Colonial Philippine Writing System) using Image Processing

Abstract

© 2020, World Academy of Research in Science and Engineering. All rights reserved. The goal of this paper is to accomplish an Optical Character Recognition (OCR) that gives an extremely contribution to the advancement of technology in terms of image recognition in Machine Learning. The researcher introduces the Feed-Forward Neural Network with Dropout Method (FFNNDM) and Convolutional Neural Network with Dropout Method (CNNDM) for the recognition of the Baybayin symbols. The phases of preprocessing of data also describe in this paper to be feed into the image recognition algorithms. The arrangement of FFNNDM is composed of one (1) dense input layer and then having a four dense (4) hidden layer and one (1) dense output layer, and the CNN structure is composed of three (3) convolutional layer, two (2) dense hidden layer and one (1) output layer. The result shows that FFNNDM is more accurate and gains an accuracy of 92.4%, loss of 0.25% and error rate of 7.55%, while the CNNDM had only an accuracy of 91.69%, loss of 0.31% and error rate of 8.31%. It also presents the confusion matrix of each algorithm to exhibit the true correct predictions of each Baybayin symbols.