PROJECT TITLE :

Adaptive Metric Learning for Saliency Detection - 2015

ABSTRACT:

During this paper, we tend to propose a unique adaptive metric learning algorithm (AML) for visual saliency detection. A key observation is that the saliency of a superpixel can be estimated by the distance from the most certain foreground and background seeds. Rather than measuring distance on the Euclidean space, we have a tendency to gift a learning method based mostly on 2 complementary Mahalanobis distance metrics: one) generic metric learning (GML) and a couple of) specific metric learning (SML). GML aims at the world distribution of the whole coaching set, whereas SML considers the specific structure of a single image. Considering that multiple similarity measures from different views could enhance the relevant info and alleviate the irrelevant one, we have a tendency to try to fuse the GML and SML along and experimentally find the combining result will work well. Different from the foremost existing strategies which are directly primarily based on low-level options, we tend to devise a superpixelwise Fisher vector coding approach to better distinguish salient objects from the background. We conjointly propose an correct seeds selection mechanism and exploit contextual and multiscale info when constructing the ultimate saliency map. Experimental results on numerous image sets show that the proposed AML performs favorably against the state-of-the-arts.