This paper demonstrates the framework and results from the team “Florida International University - University of Miami (FIU-UM)” in TRECVID 2014 Semantic Indexing (SIN) task [Smeaton09]. Two runs were submitted, and the summary of these four runs are as follows: 2B_M_A_FIU-UM.14_3: RSPM (Collateral Representative Subspace Projection Modeling) - RSPM based ranking using key frame (KF) features. 2B_M_A_FIU-UM.14_4: RSPM - RSPM based ranking using KF features. In Runs 1 and 2, the same baseline RSPM-based model is applied. We utilized the approach that was used in the 2012 SIN task to see if we could continue to improve the gains. However as a result, the performance was poor and provided no discernible gains.
