Local Interaction and Global Guidance Based Low and High-level Feature Fusion for RGB-D Fixation Prediction
Xinyue Zhang, Ting Jin, Mingjie Han, Jingsheng Lei · 2021
Saliency prediction on RGB-D images is an underexplored and challenging task in computer vision. We propose a channel-wise attention and contextual interaction asymmetric network for RGB-D saliency prediction. In the proposed network, a common feature extractor provides cross-modal complementarity between the RGB image and the corresponding depth map. In addition, we introduce a four-stream feature-interaction module that fully leverages multiscale and cross-modal features for extracting contextual information. Moreover, we propose a channel-wise attention module to highlight the feature representation of salient regions. Finally, we refine coarse maps through a corresponding refinement block. Experimental results show that the proposed network achieves performance comparable with state-of-the-art saliency prediction methods on two representative datasets.