Hierarchical Multimodal Adaptive Fusion (HMAF) Network for Prediction of RGB-D Saliency

Figure 3

Two-input attention module (attention modules 1, 2, and 3).