SAIVT-ADMRG @ MediaEval 2014 Social Event Detection

摘要: This paper outlines the approach taken by Speech, Audio, Image and Video Technologies laboratory, Applied Data Mining Research Group (SAIVT-ADMRG) in 2014 MediaEval Social Event Detection (SED) task. We participated event based clustering subtask (subtask 1), focused on investigating incorporation of image features as another source data to aid clustering. In particular, we developed a descriptor around use super-pixel segmentation, that allows low dimensional feature incorporates both colour texture information be extracted used within popular bag-of-visual-words (BoVW) approach.