Transfer Learning of a Deep Learning Model for Exploring Tourists’ Urban Image Using Geotagged Photos

Recently, as computer vision and image processing technologies have rapidly advanced in the artificial intelligence (AI) field, deep learning been applied field of urban regional study through transfer learning. In tourism studies are emerging to analyze tourists’ by identifying visual content photos. However, previous limitations properly reflecting unique landscape, cultural characteristics, traditional elements region that prominent tourism. With purpose going beyond these studies, we crawled 168,216 Flickr photos, created 75 scenes 13 categories a tourist’ photo classification analyzing characteristics photos posted tourists developed model continuously re-training Inception-v3 model. The final shows high accuracy 85.77% for Top 1 95.69% 5. was entire dataset regions attraction Seoul. We found feel attracted Seoul where modern features such skyscrapers uniquely designed architectures palaces mixed together city. This work demonstrates tourist suitable local process effectively classify large volume