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

Figure 4

Three-input attention module (attention module 4).