Abstract
Given a set of geotagged photos with a view direction, the discovering real regions of interest (DRRI) problem identifies real regions of Interest (RealROIs) in the photos. The problem is important in many societal applications, including tourist route recommendations and travel advertisements. Much work utilized geotagged photos to discover regions of interest or routes and to provide recommendations according to similar preferences. However, these approaches are not ideal for identifying RealROIs because the usage of geotagged photos is limited to the GPS coordinates of the photo. To remedy this issue, we propose the DRRI problem that can use geometric and directional information to identify the interesting regions and develop novel algorithms to discover RealROIs in photos based on the direction in which the photos were taken. Our experimental results and case studies show that our approach outperforms the related work.
Funding Information
  • National Research Foundation of Korea (NRF) grant
  • Korean Government through the Ministry of Science and ICT (2020R1F1A1048432)
  • Korea Agency for Infrastructure Technology Advancement (KAIA) grant
  • Ministry of Land, Infrastructure and Transport (RS-2022-00143336)
  • U.S. National Science Foundation (1844565)

This publication has 19 references indexed in Scilit: