Web Image Clustering Based on Multi-instance

In image retrieval and annotation, multi-instance learning has been studied actively. Most of the methods solve the MIL problem in a supervised way. In this paper, we proposed two unsupervised frameworks for clustering multi-instance objects based on expectation maximization (EM) and iterative heuristic optimization respectively. For each framework, we introduced three new algorithms of finding users´ interests on specific Web images without any manual labeled data. And comparative studies have shown the effectiveness of the proposed algorithms.