Fusing features from different sources is a critical aspect of many computer vision tasks. Existing approaches can be roughly categorized as parameter-free or learnable operations. However, ...