Abstract

Low-level saliency cues or priors do not produce goodenough saliency detection results especially when thesalient object presents in a low-contrast background with confusing visual appearance. This issue raises a seriousproblem for conventional approaches. In this paper, wetackle this problem by proposing a multi-context deep learn-ing framework for salient object detection. We employ deepConvolutional Neural Networks to model saliency of objects in images. Global context and local context are both taken into account, and are jointly modeled in a unified multicontext deep learning framework. To provide a better initialization for training the deepneural networks, we investigate different pre-training strategies, and a task-specific pre-training scheme is designed to make the multi-context modeling suited for saliency detection. Furthermore, recently proposed contemporary deep models in the ImageNet Image Classification Challenge are tested, and their effectiveness in saliency detection are investigated. Our approach is extensively evaluated on five public datasets, and experimental results show significant and consistent improvements over the state-ofthe-art methods.