Tourist Attractions Classification using ResNet

Abstract

Smart tourism is a keyword for describe the tourist on emerging forms of ICT. One application of smart tourism is to classify tourist attractions automatically, where the data in the form of pictures taken by tourists. However, there are some problems in application of tourist attractions classifications. First, in one place may have different objects and traits. Second, in some places may have a similar architecture, so it could be difficult for the system to classify the places. In this study, we focused on the tourist attractions in Jakarta and Depok using ResNet50. We divided this study into 2 scenarios. Scenario 1 is a model with 12 classes, and scenario 2 is a model with 16 classes. The results are ResNet50 has been able to handle both research problems, although not yet maximized, with average accuracy in scenario 1 is 92.17% and scenario 2 is 93.75%.