This is published under the terms of the Creative Commons Attribution licence.

Downloaded: 7558 times

Abstract

The advent of mobile devices and media cloud services has led to the unprecedented growth of personal photo collections. One of the fundamental problems in managing the increasing number of photos is automatic image tagging. Image tagging is the task of assigning human-friendly tags to an image so that the semantic tags can better reflect the content of the image and therefore can help users better access that image. The quality of image tagging depends on the quality of concept modeling which builds a mapping from concepts to visual images. While significant progresses are made in the past decade on image tagging, the previous approaches can only achieve limited success due to the limited concept representation ability from hand-crafted features (e.g., Scale-Invariant Feature Transform, GIST, Histogram of Oriented Gradients, etc.). Further progresses are made, since the efficient and effective deep learning algorithms have been developed. The purpose of this paper is to categorize and evaluate different image tagging approaches based on deep learning techniques. We also discuss the relevant problems and applications to image tagging, including data collection, evaluation metrics, and existing commercial systems. We conclude the advantages of different image tagging paradigms and propose several promising research directions for future works.