JCSE, vol. 13, no. 3, pp.89-98, September, 2019

DOI: http://dx.doi.org/10.5626/JCSE.2019.13.3.89

Deep-Learning Seat Selection on a Tour Bus Based on Scenery and Sunlight Information

Ki Hong Kim and Kwanyong Lee
Department of Computer Science, Graduate School, Korea National Open University, Seoul, Korea

Abstract: When traveling on a tour bus, the seat one chooses for viewing scenery is one of the main factors affecting one's enjoyment of a trip. However, such scenery information is not available in advance. Therefore, it is necessary to predict the scenery for a tour bus route. In previous research, such predictions have been attempted through machine learning. However, the prediction result has only informed users about which direction is best, not about how good that direction is. Moreover, no information was given about sunlight, which can also affect the viewing of scenery. Therefore, in this paper, we propose the Beautiful Scenery & Cool Shade system that quantifies the information about scenery and sunlight in four directions using deep learning and the azimuth theory. More specifically, we used ResNet-152, DenseNet-161, and Inception v3 for the prediction, and we used Google Street View for the input data. After building the system, we tested its applications to two existing tour bus routes. The results showed that our system outperformed the previous system. The proposed system allows tourists to make satisfactory travel plans and allows tour companies to develop more valuable tour services, ultimately contributing to the development of the global tourism industry.

Keyword: Deep learning; Transfer learning; Google Street View; Tour

Full Paper:   375 Downloads, 1612 View