Leveraging geo-referenced digital photographs

• Authors: Hector Garcia-Molina;Mor Naaman
• Venue: Leveraging geo-referenced digital photographs
• Year: 2005

Given automatically captured metadata such as time and location about photos in a personal collection, we devised a series of methods for supporting photo management. These methods allow an enhanced level of semantic interaction with photo collections, while requiring little or no effort from the collection's owner. This work describes how we automatically organize such geo-referenced collections into semantic location and event hierarchies that can greatly improve the user's browse and search tasks. We present an evaluation of our browsing system, PhotoCompas, including a comparison to a map-based photo browsing approach. We also show how browsing geo-referenced collections can be augmented with other types of context that is derived from time and location and is useful for retrieval in the context of personal photo collections. In addition, we show two ways in which time and location data, coupled with a minimal amount of user annotation, can effectively suggest some of the semantic content of non-annotated photographs. First, as the user annotates some of the identities of people in their collection, patterns of re-occurrence and co-occurrence of different people in different locations and events emerge. Our system uses these patterns to generate label suggestions for identities that were not yet annotated, effectively guessing the “content” of photos in terms of the people that appear in them. Second, we describe LOCALE, a system that allows users to implicitly share labels for photographs based on location. For a photograph with no label, LOCALE can assign a tentative label using knowledge about other photographs that were taken in the same area. The system thus allows automated label suggestions and text-based search for unlabeled photos. LOCALE effectively guesses the “content” of photos in terms of landmarks that appear in them.