A Deep Learning Method for Salient Object Detection

Abstract

AbstractSalient object detection is a traditional topic in computer vision that has attracted a lot of interest recent years. It detects the most salient object in an image and then segments the whole region of the salient object from the image. Since salient object detection can extract salient foreground object from the background, it is useful for several applications, such like image segmentation, object recognition, and target tracking. In this paper, we propose a novel deep learning method for salient detection based on Fully Convolutional Neural Networks (FCNs). The proposed scheme merges different level of features: to combines the side outputs of different layers to get both details and region information and generates global features and local features by using different size of pixel matrices. We tested our method on several different widely used benchmark data sets and received state-of-the-art results by comparing other methods with advantages of high efficiency and high accuracy rate.KeywordsSalient object detectionFcnsGlobal featuresLocal featuresPixel matrices