Related Experiment Video
Updated: Jun 12, 2026
12:49
A Method for 3D Reconstruction and Virtual Reality Analysis of Glial and Neuronal Cells
Published on: September 28, 2019
12.6K
Enhanced Atrous Spatial Pyramid Pooling Feature Fusion for Small Ship Instance Segmentation
Rabi Sharma1, Muhammad Saqib1,2, C T Lin1
1School of Computer Science, University of Technology Sydney, Broadway, Sydney 2007, Australia.
Journal of Imaging
|December 27, 2024
Summary
This study introduces an enhanced Atrous Spatial Pyramid Pooling (ASPP) feature fusion method to improve the instance segmentation of small ships in maritime settings. The novel approach significantly boosts accuracy where other algorithms fail.
Area of Science:
- Computer Vision
- Maritime Surveillance
- Deep Learning
Background:
- Accurate instance segmentation of small ships in maritime environments is critical for applications like navigation safety and security.
- Existing algorithms struggle to detect and segment small ships due to their limited appearance, small size, and distant locations.
Purpose of the Study:
- To develop a novel method for enhancing the instance segmentation of small ships.
- To address the limitations of current algorithms in detecting small maritime objects.
Main Methods:
- Proposed an enhanced Atrous Spatial Pyramid Pooling (ASPP) feature fusion module specifically designed to refine and fuse features of small objects.
- Integrated the enhanced ASPP module into an instance segmentation framework.
Main Results:
- The proposed framework demonstrated superior performance compared to state-of-the-art models like Mask R-CNN and SOLOv2.
- Achieved high average precision (mask AP) scores: 75.8% on ShipSG, 69.5% on ShipInsSeg, and 54.5% on MariBoats datasets.
Conclusions:
- The enhanced ASPP feature fusion method effectively improves small ship instance segmentation in maritime scenes.
- The developed framework offers a robust solution for accurate detection and segmentation of small vessels, outperforming existing methods.
