Deep Learning for Malay Architectural Identification: A CNN Approach to Heritage Recognition and Preservation

Abstract

This study develops a classification model of traditional Malay buildings using a deep learning approach to analyze the suitability between the design model and real objects. We utilized VGG16 with a dataset of Malay traditional building images to train and test the model. The test results show that the VGG16 model can achieve an accuracy of 98.77% with a learning rate of 0.0001, dropout of 0.20, and epochs of 25. These results indicate that VGG16 is effective as a tool in the process of identifying and preserving traditional architecture based on imagery by producing good accuracy.