Classifying Architectural Images of Digital Heritage: A CNN-SVM Hybrid Approach

India is a nation blessed with a plethora of cultural landmarks, including 37 UNESCO World Heritage Sites, many of which are world-famous architectural structures. We must protect cultural heritages because they bind successive generations together throughout time. Researchers, travelers, architects, etc. They visit several historical locations, where it is frequently challenging for them to recognize and learn more about the historical significance of the building they are fascinated by. Due to the scope and accuracy of the information, it is difficult to archive, preserve, and share knowledge about these cultural treasures. It is easier to examine and comprehend heritage assets when images are accurately predicted to correspond to their right name (heritage site). It takes an enormous amount of time and work to classify data that includes images. In this study, the authors have designed a new hybrid model used to extract the features from architectural images and further classified into different shapes of the architectural sites and achieved 95% accuracy and the highest precision rate is achieved for the Altar class i.e. 88.05%, the highest recall rate is achieved for the Vault class i.e. 87.36
^%
, and the highest Fl score rate is achieved for Flying buttress i.e. 89.10%.