Fig. 5: The proposed attention mechanism in the multi-scale feature fusion module.
From: A deep learning based automatic report generator for retinal optical coherence tomography images
The semantic information of deep features was used to guide the model to focus on regions of interest.
