A new framework for high-level feature extraction
Zan Gao, Xiaoming Nan, Tao Liu, Zhicheng Zhao, Anni Cai · 2009
A new framework for high-level feature extraction (or semantic concept detection) is proposed. In this system, features at different granularities are extracted, and four classifiers with complementary features for each concept are employed, and then the results are fused. We have evaluated 18 fusion schemes, and choose the best one for each concept to form the final results. The experiments on the auto-test corpus and TRECVID-2008 corpus show that the proposed system is effective and stable.