Local Context Attention for Salient Object Segmentation

Salient object segmentation aims at distinguishing various salient objects from backgrounds. Despite the lack of semantic consistency, often have obvious texture and location characteristics in local area. Based on this priori, we propose a novel Local Context Attention Network (LCANet) to generate locally reinforcement feature maps uniform representational architecture. The proposed network introduces an Attentional Correlation Filter (ACF) module explicit attention by calculating correlation map between coarse prediction global context. Then it is expanded Block (LCB). Furthermore, one-stage coarse-to-fine structure implemented based LCB adaptively enhance context description ability. Comprehensive experiments are conducted several datasets, demonstrating superior performance LCANet against state-of-the-art methods, especially with 0.883 max F-score 0.034 MAE DUTS-TE dataset.