Impact of Number of Clusters in Video Search Reranking by Fusion
Arambam Neelima, K. Veningston · CiiT international journal of data mining and knowledge engineering · 2011
Reranking techniques for video search results has been an important research topic over the years. Many methods have been proposed considering both the overall and the topmost result performance. One of such method is done by fusing multiple modalities in different feature spaces. This paper presents the impact of number of clusters on such a reranking scheme. Experiments show that this will largely affect the final reranking performance. For clustering NCut clustering algorithm has been used which is a spectral clustering algorithm.