A review of image set classification

In computer vision, we generally solve a classification problem by a single image. With the video cameras being widely used in our real life, it is a nature choice to solve a classification problem by image sets. Compared with the single image based methods, the image set classification deals with severe changes of appearance and makes decisions by comparing the query set with gallery sets. So the image set classification offers more promises and has therefore attracted significant research attention in recent years. In this paper, we provide a review on image set classification. Our review begins with an overview of the direction of image set classification. Then we detail some classic algorithms. Experimental analyses are provided in corresponding subsection to compare classification performance of various methods and draw some meaningful conclusions. Finally, several promising directions and tasks are provided as guidelines for future work. (c) 2018 Elsevier B.V. All rights reserved.