In this paper, we propose an end-to-end group-wise deep co-saliency detectionapproach to address the co-salient object discovery problem based on the fullyconvolutional network (FCN) with group input and group output. The proposedapproach captures the group-wise interaction information for group images bylearning a semantics-aware image representation based on a convolutional neuralnetwork, which adaptively learns the group-wise features for co-saliencydetection. Furthermore, the proposed approach discovers the collaborative andinteractive relationships between group-wise feature representation andsingle-image individual feature representation, and model this in acollaborative learning framework. Finally, we set up a unified end-to-end deeplearning scheme to jointly optimize the process of group-wise featurerepresentation learning and the collaborative learning, leading to morereliable and robust co-saliency detection results. Experimental resultsdemonstrate the effectiveness of our approach in comparison with thestate-of-the-art approaches.
展开▼
