DeepDBSCAN: Deep Density-Based Clustering for Geo-Tagged Photos

Density-based clustering algorithms have been the most commonly used for discovering regions and points of interest in cities using global positioning system (GPS) information geo-tagged photos. However, users sometimes find more specific areas real objects captured pictures. Recent advances deep learning technology make it possible to recognize these since detection is a very time-consuming task, simply combining with density-based costly. In this paper, we propose novel algorithm supporting content clustering, called spatial applications noise (DeepDBSCAN). DeepDBSCAN incorporates object by into density nearest neighbor graph technique. Additionally, supports graph-based reduction that reduces number detections. We performed experiments pictures shared on Flickr compared performance multiple demonstrate excellence proposed algorithm.