UEC at TRECVID 2012 SIN and MED task

In this paper, we describe our approach and results for the semantics indexing (SIN) task and Multimedia event detection (MED) task at TRECVID2012. In our run of SIN task, we used three features, spatio-temporal (ST) features, SURF and color features. This year, we use all frame to extract features. This run used Multiple Kernel Learning as a fusion method to combine all these features in the same way as last year. Our submitted run is F A UEC1 1. As a result of the full-category SIN task, run reached a performance infAP=0.116. In MED task, we divide videos to shots which are 3000 frames at most and extract SURF, ST features from shots. Then, we select positive shots with VisualRank method from. We get the average of the top three shot scores as the original video score.