No abstract

In this paper, we describe a retrieval system that uses hidden annotation to improve the performance. The contribution ofthis paper is a novel active learning framework that can improve the annotation efficiency. For each object in the database, we maintain a list of probabilities, each indicating the probability of this object having one of the attributes. This list of probabilities serves as the basis of our active learning algorithm, as well as semantic features to determine the similarity between objects in the database. We show active learning has better performance than random sampling in all our experiments.

No abstract

No abstract

Most of the existing work on modeling variable bit rate (VBR) video sources does not either explicitly take into account groupof-pictures (GOP) or assumes a fixed GOP structure. Real video data inherently possesses a variable GOP structure. We propose a number of doubly Markov models for such real data. These models outperform presently proposed models and have reasonable complexity in terms of the number of parameters.