Hierarchical Feature Fusion

Hierarchical Feature Fusion (HFF): An Effective Method for Image Model Blocks What is Hierarchical Feature Fusion? Hierarchical Feature Fusion (HFF) is a method of fusing feature maps obtained by convolving an image with different dilation rates. It is used in image model blocks like ESP and EESP to eliminate unwanted artifacts caused by a large receptive field introduced by dilated convolutions. How does Hierarchical Feature Fusion work? The ESP (Efficient Spatial Pyramid) module uses dil