nLow-level features
(color, edge, texture…):
ntend to be
unstable and unreliable in representing semantics
nAlso, hard to
relate a query to suitable low level features
nFor example: “Find shots with people holding banner”
nHigh-level
features
nDetectors of varying accuracy
are available. For example, face, car, boat, commercial, sports, etc
nProvide partial semantics to ASR and queries
nHave been shown to be
effective in TRECVID 2005
nCan be easily incorporated
into text-based retrieval systems