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

Abstract

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

Improved Gated Recurrent Units together with Fusion for Semantic Segmentation of Remote Sensing Images based on Parallel Hybrid Network

“…To compensate for the lost detailed information in high-level features, Gated Fully Fusion (GFF) [28] is used to selectively fuse features from multiple levels using gates in a fully connected way. To collect contextual information and produce relevant global information, Mahmood et al [29] modify SegNet architecture to embed Gated Recurrent Units (GRU) within the convolution layers. Meanwhile, multiple CRF-GRU layers [30] are injected into an FCN to model hierarchical contexts and show competitive results for semantic segmentation.…”