Top down saliency estimation via superpixel-based discriminative dictionaries
In Proceedings British Machine Vision Conference 2014
AbstractPredicting where humans look in images has gained significant popularity in recent years. In this work, we present a novel method for learning top-down visual saliency, which is well-suited to locate objects of interest in complex scenes. During training, we jointly learn a superpixel based class-specific dictionary and a Conditional Random Field (CRF). While using such a discriminative dictionary helps to distinguish target objects from the background, performing the computations at the superpixel level allows us to improve accuracy of object localizations. Experimental results on the Graz-02 and PASCAL VOC 2007 datasets show that the proposed approach is able to achieve state-of-the-art results and provides much better saliency maps.
FilesExtended Abstract (PDF, 1 page, 240K)
Paper (PDF, 12 pages, 1.2M)