Co-Saliency Detection With Co-Attention Fully Convolutional Network

Co-saliency detection aims to detect common salient objects from a group of relevant images. Some attempts have been made with the Fully Convolutional Network (FCN) framework and achieve satisfactory results. However, due stacking convolution layers pooling operation, boundary details tend be lost. In addition, existing models often utilize extracted features without discrimination, leading redundancy in representation since actually not all are helpful final prediction some even bring distraction. this paper, we propose co-attention module embedded FCN framework, called as Co-Attention (CA-FCN). Specifically, is plugged into high-level FCN, which can assign larger attention weights on smaller ones background uncommon distractors boost performance. Extensive experiments three popular co-saliency benchmark datasets demonstrate superiority proposed CA-FCN, outperforms state-of-the-arts most cases. Besides, effectiveness our new also validated ablation studies.