Mining Weakly Labeled Web Facial Images for Search-Based Face Annotation Using Neural Network Classifier

Abstract

This paper inspects a structure for search-based-face annotation using mining weakly labeled web facial images. Facial photographs are candidly present on the Internet, from that a few facial photographs are properly labeled, however, some of them are not correctly labeled. These facial photographs are repeatedly incomplete and noisy. For enhancing tag quality of weakly net facial photographs, ULR approach is also advantageous for cleansing or filtering the tags of net facial photographs (Wang et al. in IEEE Trans Knowl Data Eng 26, [1]). Big headache issue for search-based face annotation scheme is, whenever the given test facial portrait is not a common person, there are no much more same facial photographs present on the web. A supervised appropriate name tag can be given to a test face portrait by employing face annotation using search-based paradigm, but it also increases the efficiency and scalability. The supervised neural network classifier approach is looking to optimize the tag quality of face portrait by majority voting against the face annotation by search-based paradigm.