A Mask Attention Interaction and Scale Enhancement Network for SAR Ship Instance Segmentation

TL;DR

This paper introduces MAI-SE-Net, a novel network for SAR ship instance segmentation that enhances multi-scale feature interaction and improves small ship detection performance.

Contribution

The paper proposes MAI-SE-Net, combining attention interaction and scale enhancement modules to improve SAR ship segmentation, especially for small ships, outperforming existing models.

Findings

Outperforms nine competitive models on SSDD and HRSID datasets.

Achieves 4.7% higher detection AP on SSDD.

Achieves 3.4% higher segmentation AP on SSDD.

Abstract

Most of existing synthetic aperture radar (SAR) ship in-stance segmentation models do not achieve mask interac-tion or offer limited interaction performance. Besides, their multi-scale ship instance segmentation performance is moderate especially for small ships. To solve these problems, we propose a mask attention interaction and scale enhancement network (MAI-SE-Net) for SAR ship instance segmentation. MAI uses an atrous spatial pyra-mid pooling (ASPP) to gain multi-resolution feature re-sponses, a non-local block (NLB) to model long-range spa-tial dependencies, and a concatenation shuffle attention block (CSAB) to improve interaction benefits. SE uses a content-aware reassembly of features block (CARAFEB) to generate an extra pyramid bottom-level to boost small ship performance, a feature balance operation (FBO) to improve scale feature description, and a global context block (GCB) to refine features. Experimental results on two public SSDD and HRSID datasets reveal that MAI-SE-Net outperforms the other nine competitive models, better than the suboptimal model by 4.7% detec-tion AP and 3.4% segmentation AP on SSDD and by 3.0% detection AP and 2.4% segmentation AP on HRSID.