Attention to the Scale: Deep Multi-Scale Salient Object Detection

Abstract

Salient object detection has been greatly boosted thanks to the deep convolutional neural networks (CNN), especially fully convolutional neural networks (FCN). Nowadays, it is possible to train an end-to-end deep model for salient object detection. However, the diverse scales of salient objects still pose major challenges for these state-of-the-art methods. In this paper, we investigate how different scales of context information affect the performance of salient object detection by building our saliency prediction model on a pyramid spatial pooling network. An attention-to-scale model is trained to measure the importance of saliency features at different scales, and a saliency fusion stage is utilized to extract complementary information from different scales. The proposed model is trained in an end-to-end manner. Extensive experimental results on eight benchmark datasets demonstrate the superior performance of our proposed method compared with existing state-of-the-art methods.