Related Experiment Video
Updated: May 28, 2026
03:31
End-To-End Deep Neural Network for Salient Object Detection in Complex Environments
Published on: December 15, 2023
Structure-Enhanced Underwater Object Detection via Wavelet-Edge Collaboration and Selective Multi-Scale Fusion
Dejun Li1, Chunrong He1, Peng Tu2
1State Key Laboratory of Deepsea Manned Vehicle, China Ship Scientific Research Center, Wuxi 214082, China.
Sensors (Basel, Switzerland)
|May 27, 2026
Summary
A new underwater object detector, WEC-UOD, enhances structural details and fuses multi-scale features to improve detection accuracy in challenging marine environments. It outperforms existing baselines on benchmark datasets.
Area of Science:
- Marine Biology
- Computer Vision
- Robotics
Background:
- Underwater object detection is crucial for ocean exploration and marine applications.
- Degraded image quality (absorption, scattering, interference) hinders detection of small or varied objects.
Purpose of the Study:
- To develop an improved underwater object detector (WEC-UOD) that enhances structural representation and feature fusion.
- To address challenges posed by poor underwater image quality without a separate enhancement step.
Main Methods:
- Proposed WEC-UOD detector with a Wavelet-Edge Collaboration (WEC) module for structural compensation and edge refinement.
- Implemented a Scale-Selective Fusion (SSF) module for adaptive feature aggregation and background suppression.
Main Results:
- WEC-UOD achieved mAP@0.5 scores of 87.4% on RUOD and 86.9% on DUO.
- Consistently outperformed the YOLOv11s baseline on both datasets.
Conclusions:
- The WEC-UOD effectively combines structural enhancement with selective multi-scale feature aggregation for superior underwater object detection.
- The proposed modules significantly improve performance in challenging underwater conditions.