1. Subspace mapping of facial attributes for face image retrieval
Authors : V.padmini, S.suchitra
Pages : 323-328
DOI : http://dx.doi.org/10.21172/1.83.048
Keywords : Face imageFacial attributesHigh level cuesSimilarity SearchImage Retrieval Abstract :In the growing advancements of recent multimedia applications, the domain of Face images and its corresponding applications have retained its great importance. The wide array of social, real-time and crime investigation applications and their expanding field of necessities demands a far more accurate and fast face image retrieval system. Most face image retrieval techniques depend on only low-level cues to compare the face images of the people based on similar their visual identity. In this paper a face image retrieval system driven by enabled facial attributes is proposed to efficiently utilise the facial attributes information as high level semantic cues to discriminate people’s identities by a semi-supervised approach of mapping the subspace labels along with the extracted low level components in the run time environment. An inverted indexing Scheme is used for the storing and retrieving face image representations. Our aim is minimise the possible marginal errors of inter-class and intra-class variations detected under different illumination conditions during the similarity search. We demonstrate relative improvement of 87% in MAP by experimenting and comparing our proposed method and other state of the art methods in LFW and PublicFig datasets.
Citing this Journal Article :V.padmini, S.suchitra, "Subspace mapping of facial attributes for face image retrieval", Volume 8 Issue 3 - May 2017, 323-328
Click here to Submit Copyright Takedown Notice for this article.