Video Annotation System Based on Categorizing and Keyword Labelling

摘要

In this work, we demonstrate an automatic video annotation system which can provide users with the representative keywords for new videos. The system explores the hierarchical concept model and multiple feature model to improve the effectiveness of annotation, which consists of two components: a SVM classifier to ascertain the category; and a multiple feature model to label the keywords. We implement the demo system using the videos downloaded from YouTube. The results show the superiority of our approach.