Automated Image Generation Reflecting Current Status of PoIs for Supporting On-Site Tourist Destination Selection

To support the decision-making process for selecting upcoming tourist spots (PoI: Points of Interest), it is necessary to share the tourist destinations' current status. Methods for sharing the current status of PoIs could include installing live cameras or having tourists share real-Time photos. Still, there are challenges in terms of installation and maintenance costs and privacy concerns. In this paper, a tourism support system is proposed to aid in the decision-making of which PoI to visit next by automatically generating PoI current status images from template images and contextual information of tourist destinations while preserving privacy and presenting them to users. In the proposed system, semantic segmentation is used to divide template images into categories such as sky class, trees/grasses class, crowd class, and overall class. Based on the separately collected contextual information (weather, condition of trees/grass, crowd density, time, etc.), appropriate image transformations (using style transfer, etc) are applied to each category. The proposed system consists of an automated method for generating PoI status images and an application that presents these images to users. As part of a tourism experiment using the proposed system, 11 individuals aged 20 to 30 participated in a roughly 2-hour tour of Nara City in Japan. The results showed that compared to presenting template images, the proposed method yielded a decision-making score improvement of 0.27 and a similarity (to the current status) score improvement of 0.45 (out of five), both of which were statistically significant.