Research And Implementation Of Personal Photo Organizer Based On Face Recognition Technology

In the past decade, most of the personal photo organization applications and photo sharing websites still used textual information for image searching. When the number of images is restricted and only few people will needed to access huge repositories of images, text-based searching was quite feasible. But in recent year, the widespread of digital cameras and the low prices of the data storage device, thousands and thousands of digital photos has been generated by individuals. A statistics number of photos host on Flikcr, which is currently the biggest image hosting and sharing website, claims of hosting more than 3 billion images at the end of 2012.As a result, more and more people find themselves having thousands of digital photos with little organization and little utility, and they are resigned to gain no more enjoyment from finding the object in this whole mass. Yet it becomes increasingly important to devise efficient and effective methods to help users to search, browse, retrieval or annotate their personal photo collections.Current research toward solving personal photo management suffer from two problems:(1) lacking of training data, and(2) no consolidated reference for classification. In this thesis, we propose an automated annotation framework to address these problems. The framework is composed by three main components: the context information generator, the semantic concept detector, and the face identification model. Context information is extracted from the exif header file; semantic concept detector is trained by the Support Vector Machine(SVM); face recognition model is a combination of face detection and recognition. By assigning multi-labels for each photo, the framework makes the photo collection more structured and search-able. Our experimental results show that the techniques used in this framework are promising.