Multi-Guidance CNNs for Salient Object Detection

Feature refinement and feature fusion are two key steps in convolutional neural networks–based salient object detection (SOD). In this article, we investigate how to utilize multiple guidance mechanisms better refine fuse extracted multi-level features propose a novel multi-guidance SOD model dubbed as MGuid-Net. Since boundary information is beneficial for locating sharpening objects, edge utilized our network together with saliency SOD. Specifically, self-guidance module applied features, respectively, which aims gradually guide the of lower-level by higher-level features. After that, cross-guidance devised mutually via complementarity between them. Moreover, integrate refined also present an accumulative module, exploits high-level different hierarchical manner. Finally, pixelwise contrast loss function adopted implicit help retain more details objects. Extensive experiments on five benchmark datasets demonstrate can identify regions image effectively compared most state-of-the-art models.