Volume 7 Issue 1 - May 2016

  • 1. Automatic annotation to rank the images

    Authors : Sunil Hebbale, Ashwini Gavali

    Pages : 301-308

    DOI : http://dx.doi.org/10.21172/1.71.043

    Keywords : Automatic image annotation, tag ranking, matrix recovery, low-rank, trace norm.

    Abstract :

    Due to the popularity of image capturing digital devices and the ease of social network/photo sharing services (e.g., Facebook, Twitter, Flickr), image annotation came into limelight due to its application in image matching and retrieval. Previously image annotation was casted into a multilabel classification problem which had a drawback that it required a large number of training images with clean and complete annotations in order to learn a reliable model for tag prediction. To overcome this limitation we develop a novel approach that combines the strength of tag ranking with the power of matrix recovery. In this work tags are ranked in the descending order of their relevance to the given image, thus simplifying the problem. The proposed method also aggregates the prediction models into a matrix, and casts tag ranking into a matrix recovery problem which introduces the matrix trace norm to explicitly control the model complexity so that tag ranking can be done even when the tag space is large and the number of training images is limited. Experiments on various image data sets show the effectiveness of the proposed framework for tag ranking compared with the previous approaches for image annotation and tag ranking.

    Citing this Journal Article :

    Sunil Hebbale, Ashwini Gavali, "Automatic annotation to rank the images", https://www.ijltet.org/journal_details.php?id=901&j_id=3059, Volume 7 Issue 1 - May 2016, 301-308, #ijltetorg