用于小船舶实例分割的增强空洞空间金字塔池化特征融合

Enhanced Atrous Spatial Pyramid Pooling Feature Fusion for Small Ship Instance Segmentation.

作者信息

Sharma Rabi, Saqib Muhammad, Lin C T, Blumenstein Michael

机构信息

School of Computer Science, University of Technology Sydney, Broadway, Sydney 2007, Australia.

National Collections & Marine Infrastructure, CSIRO, Sydney 2007, Australia.

出版信息

J Imaging. 2024 Nov 21;10(12):299. doi: 10.3390/jimaging10120299.

DOI:10.3390/jimaging10120299

PMID:39728196

原文链接:https://pmc.ncbi.nlm.nih.gov/articles/PMC11677345/

Abstract

In the maritime environment, the instance segmentation of small ships is crucial. Small ships are characterized by their limited appearance, smaller size, and ships in distant locations in marine scenes. However, existing instance segmentation algorithms do not detect and segment them, resulting in inaccurate ship segmentation. To address this, we propose a novel solution called enhanced Atrous Spatial Pyramid Pooling (ASPP) feature fusion for small ship instance segmentation. The enhanced ASPP feature fusion module focuses on small objects by refining them and fusing important features. The framework consistently outperforms state-of-the-art models, including Mask R-CNN, Cascade Mask R-CNN, YOLACT, SOLO, and SOLOv2, in three diverse datasets, achieving an average precision (mask AP) score of 75.8% for ShipSG, 69.5% for ShipInsSeg, and 54.5% for the MariBoats datasets.

摘要

在海洋环境中，小船的实例分割至关重要。小船的特点是外观有限、尺寸较小且在海洋场景中位于远处。然而，现有的实例分割算法无法检测和分割它们，导致船舶分割不准确。为了解决这个问题，我们提出了一种名为增强空洞空间金字塔池化（ASPP）特征融合的新颖解决方案，用于小船实例分割。增强的ASPP特征融合模块通过细化小目标并融合重要特征来聚焦于小目标。该框架在三个不同的数据集中始终优于包括Mask R-CNN、Cascade Mask R-CNN、YOLACT、SOLO和SOLOv2在内的现有最先进模型，在ShipSG数据集上的平均精度（掩码AP）得分达到75.8%，在ShipInsSeg数据集上为69.5%，在MariBoats数据集上为54.5%。