Video Concept Detection Using Support Vector Machines – TRECVID 2007 Evaluations

This report describes video concept detection using Support Vector Machine (SVM) over TRECVID 2007 corpus. We perform the experiments on low-level features extraction, data preparation and classification procedure. Through analyzing the characteristics of the TRECVID 2007 data set, we mainly focus on data preparation for training concept detectors, as well as the preparation of auxiliary training data by using TRECVID 2005 data.