A Novel Approach to Search Based Face Annotation (SBFA) Using Weakly Labeled Web Facial Images

Abstract

This paper introduces a framework of search based face annotation (SBFA) by mining weakly labeled web facial image that are available on www. one challenging part of SBFA scheme is managing most similar web facial images and their weak labels. To avoid this problem, I propose an effective refinement of unsupervised label(RUL) scheme for refining the labels of web facial images. To speed up the proposed system, I also propose a clustering-based approximation algorithm which can improve the scalability.