Attention and boundary guided salient object detection

In recent years, fully convolutional neural network (FCN) has broken all records in various vision task. It also achieves great performance in salient object detection. However, most of the state-of-the-art methods have suffered from the challenge of precisely segmenting the entire salient object with uniform region and explicit boundary and effectively suppressing the backgrounds on complex images. There is still a large room for improvement over the FCN-based saliency detection approaches. In this paper, we propose an attention and boundary guided deep neural network for salient object detection to better locate and segment the salient objects with uniform interior and explicit boundary. A channel-wise attention module is utilized to emphasize the important regions, which selects the important feature channels and assigns large weights to them. A boundary information localization module is proposed for suppressing the irrelevant boundary information to better locate and explore the useful structure of objects. The proposed approach achieves state-of-the-art performance on four well-known benchmark datasets. (C) 2020 Elsevier Ltd. All rights reserved.