Employing GRU to combine feature maps in DeeplabV3 for a better segmentation model

In this paper, we aim to enhance the segmentation capabilities of DeeplabV3 by employing Gated Recurrent Neural Network (GRU). A 1-by-1 convolution in was replaced GRU after Atrous Spatial Pyramid Pooling (ASSP) layer combine input feature maps. The and have sharable parameters, though, latter has gates that enable/disable contribution each map. experiments on unseen test sets demonstrate instead would produce better results. used datasets are public provided MedAI competition.