Conference/ProceedingsACM International Conference on Multimedia Retrieval (ICMR)
Start date17.04.2011
End date20.04.2011
Author(s)Pascal Kelm, Sebastian Schmiedeke, Thomas Sikora
TitleMulti-modal, Multi-resource Methods for Placing Flickr Videos on the Map
AbstractWe present three approaches for placing videos in Flickr on
the world map. The toponym extraction and geo lookup ap-
proach makes use of external resources to identify toponyms
in the metadata and associate them with geo-coordinates.
The metadata-based region model approach uses a k-nearest-
neighbour classi´Čüer trained over geographical regions. Videos
are represented using their metadata in a text space with re-
duced dimensionality. The visual region model approach uses
a support vector machine also trained over geographical re-
gions. Videos are represented using low-level feature vectors
from multiple key frames. Voting methods are used to form
a single decision for each video. We compare the approaches
experimentally, highlighting the importance of using appro-
priate metadata features and suitable regions as the basis of
the region model. The best performance is achieved by the
geo-lookup approach used with fallback to the visual region
model when the video metadata contains no toponym.
Key wordscompression, geo-localization, gazetteers, probabilistic latent semantic analysis, MPEG-7 visual features