Abstract: A challenge in this system is to perform glossary by exploring the list of comparable facial photos and their weak tag that are much noisy and deficient. On face model, we are using unsupervised face alignment into the Lucas-Kanade image registration approach . We are using efficient optimization technique to handle appearance variations. The method is full automatic and can corpe with pose variations and expressions in photos, all in an unsupervised manner. Experiments on a more number of images showed that the approach is efficient.We are using search-based face annotation (SBFA) by mining weak labeled facial photos are freely available on the World Wide Web (WWW). Challenging problem for search-based face annotation technique is how to efficiently perform annotation by the index of similar facial photoss and their no correct labels that are noisy or incomplete. We are developing powerfull optimization techniques to solve the large-scale learning task efficiently. To increase the speed of the proposed system, we also based a clustering-based algorithm this can improve the scalability and performance. We have conducted an extensive set of empirical studies on a large-scale web facial image testbed, in which encouraging results showed that the developed ULR algorithms can boost the performance of the SBFA scheme.
Keywords: Face annotation, content-based image retrieval, machine learning, label refinement, search-based face annotation, weak label.