Multi-feature aggregation network for salient object detection

As the network deepens, the boundary information and small object will be lost, which result in blurred edges and incomplete salient detection. In this paper, we propose a multi-feature aggregated network to solve this problem. First, the edge features are gained through the shallow network. Then, we extract multi-level salient features of the encoder layer by the multi-level features extraction module. In order to make the most of these features, we combine the edge features with the multi-level features. Finally, the fused features are sent to the residual refinement module to obtain the final saliency map. The experimental results show that the proposed method performs better than other six methods on four public datasets.