Retrieving Geo-Location of Videos with a Divide & Conquer Hierarchical Multimodal Approach
Résumé
Geotagging is the process of automatically adding geographical identification metadata to media objects, in particular to images and videos. This geo-information is called geotag(s), and usually consists of the latitude and longitude world-map coordinates. Determining the place where the content has been captured dramatically extends the knowledge around the media object, especially when combined with time information. Linking time- and geographical-related content offers a new and practical way of automatically searching, organizing or visualizing personal and professional media databases. It also enables the retrieval of various relevant content such as web pages, hence providing users with a wide variety of location-specific information. In recent years, we have witnessed a dramatic increase in the number of such geotagged media data. Due to the massive spreading of GPS-enabled cameras and mobile phones, geographic coordinates are captured and attached to the content generated by these devices. However, most of the media available remain deprived of this information. For this reason, the problem of automatically assigning geotags to media content (and conversely) is a challenge that increasingly interests the Multimedia community, as reflected by the success of the Mediaeval benchmark's placing task.
Fichier principal
trevisiol_icmr.pdf (685.02 Ko)
Télécharger le fichier
icmr_70.png (9.66 Ko)
Télécharger le fichier
Origine | Fichiers produits par l'(les) auteur(s) |
---|
Format | Figure, Image |
---|
Loading...