Edge-Guided Non-Local Fully Convolutional Network for Salient Object Detection

Fully Convolutional Neural Network (FCN) has been widely applied to salient object detection recently by virtue of high-level semantic feature extraction, but existing FCN-based methods still suffer from continuous striding and pooling operations leading loss spatial structure blurred edges. To maintain the clear edge objects, we propose a novel Edge-guided Non-local FCN (ENFNet) perform edge-guided learning for accurate detection. In specific, extract hierarchical global local information in incorporate non-local features effective representations. preserve good boundaries guidance block embed prior knowledge into maps. The not only performs feature-wise manipulation also spatial-wise transformation embeddings. Our model is trained on MSRA-B dataset tested five popular benchmark datasets. Comparing with state-of-the-art methods, proposed method performance well