eduzhai > Applied Sciences > Engineering >

Deep Learning Based Single Sample Per Person Face Recognition A Survey

  • Save

... pages left unread,continue reading

Document pages: 8 pages

Abstract: Face recognition has been an active research area in the field of patternrecognition, especially since the rise of deep learning in recent years.However, in some practical situations, each identity in the training set hasonly a single sample. This type of situation is called Single Sample Per Person(SSPP), which brings a great challenge to the effective training of deepmodels. To resolve this issue, and to unleash the full potential of deeplearning, many deep learning based SSPP face recognition methods have beenproposed in recent years. There have been several comprehensive surveys fortraditional methods based SSPP face recognition approaches, but emerging deeplearning based methods are rarely involved. In this paper, we focus on thosedeep methods, classifying them as virtual sample methods and generic learningmethods. In virtual sample methods, virtual face images or virtual facefeatures are generated to benefit the training of the deep model. In genericlearning methods, additional multi-sample generic set are used. Efforts oftraditional methods and deep feature combining, loss function improving andnetwork structure improving are involved in our analysis in the genericlearning methods section. Finally, we discuss existing problems of identityinformation retention in virtual sample methods and domain adaption in genericlearning methods. Further, we regard the semantic gap as an important futureissue that needs to be considered in deep SSPP methods.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×