Interactive object retrieval via cost-minimizing queries

Studies using machine learning algorithms have documented advantage in applications of visual object retrieval. While interactive techniques have also been reported, the relevance feedback by minimizing the costs of time and annotation has not been investigated due to the requirement of data storage and online toolbox. In this work, we present an interactive scheme of object retrieval coupled with cost-minimizing queries in order to provide accurate results in a short time. We investigate a fast mechanism of relevance feedback to reduce time expense and explore the most reliable annotator in crowds to control annotation costs. Our experimental results demonstrate the effectiveness and efficiency of proposed framework.