Video2GPS: Geotagging using collaborative systems, textual and visual features
Pascal Kelm, Sebastian Schmiedeke, Thomas Sikora · 2010
Assigning geographical coordinates to shared content has become a popular activity on the Web, but nevertheless there are still huge amounts of media data without any geographical tags. Our approach enables these media data to be geotagged with the help of recently tagged media and knowledge-based collaborative systems. It includes three dierent methods{querying collaborative systems, document indexing, and classication based on visual features {which are combined to estimate geographical regions. Our approach can be applied under restrictions encountered in real applications. We give ve experimental results for the MediaEval 2010 placing task collection.