A Hybrid Deep Learning Approach for Multi-Classification of Heritage Monuments Using a Real-Phase Image Dataset
Vinay Kukreja, Rishabh Sharma, Satvik Vats
Abstract
The cultural heritage of a nation is defined by its historical monuments which depict its art, legacies, historical significances, tradition, culture, and teachings. The importance and value of heritage monuments are not limited to the present era but are also connected with its ancient times and period where one can able to link with its past histories. In the field of automation, Artificial Intelligence (AI) has expanded significantly in recent years as a result of technical advancement. It is now required to utilize and integrate technology in the field of digital heritage and monuments so that it may be used for conserving, preserving, restoring, and raising awareness of heritage monuments. To fill this gap, the proposed study has focused on the multi-classification of Indian monuments using hybrid convolutional neural networks (CNNs) and long short-term memory (LSTM) model for recognition and then further the multi-classification of various famous Indian heritage monuments. The complete study has been divided into two major steps covering binary classification of Indian monuments for recognition of heritage Indian monuments which yields an accuracy of 92.37%, along with this multi-classification of Indian monuments into four further classes resulting into an accuracy of 95.89%. The exceptional accuracy results for both types of classification have proved the efficiency of the DL model for image-based Indian monument classification tasks. In addition to that the proposed work assists in better quality of life, community enhancement, and technology transfer for the readers and experimenters in the aforementioned domain.