<!DOCTYPE videoTopic SYSTEM "videoTopic.dtd">
<videoTopic num="043"  interactive="N-I" automatic="Y-A" knownItems="N-K">
  <!-- Landscape extraction -->
  <textDescription text="All shots should be extracted which contain
the specific landscape: grassland. It is important not to assume any
available information based on example images. Of course, each
contestant can use a specific collection of reference images to train
algorithms for finding relevant shots."/>
  <!-- Retrieval will be based mainly on image content, audio cues
such as music or a voice-over might help in some cases. Of course
scripts can give important information whether shorts with landscapes
occur in the video. In this case the retrieval should concentrate on
combining the sequence information in the scripts with the actual video
contents.-->
  <videoExample src="BOR10.MPG" start="5.6s" stop="9.2s"/>
  <!-- grassland: from frame 168 to 276 -->
</videoTopic>
