Skip to Main Content
This paper presents a Multiple-Instance Decision Based Neural Networks (MI-DBNN) based image retrieval system. Without precisely image segmentation, the image retrieval problem is considered as a Multiple-Instance Learning problem. A set of exemplar images are selected, each of which is labelled as conceptual related (positive) or conceptual unrelated (negative) image. Then, the MI-DBNN is trained to learn the user's preferred image concept from the positive and negative examples. The proposed system is built and located on http://188.8.131.52/MIL/. Experimental results show that our method can significantly improve the retrieving performance from 68.4% to 80.7%, which outperforms to the results of some leading image retrieval methods.