A Comprehensive Study On Underwater Object Detection Using Deep Neural Networks

The escalating crisis of marine pollution, especially the buildup of underwater waste, poses a significant threat to ocean ecosystems and marine life. Identifying and removing submerged debris is challenging due to poor visibility, complex water conditions, and the inefficiency of traditional manual...

Full description

Saved in:
Bibliographic Details
Main Authors: K. Samanth, Ramyashree Ramyashree, B. N. Anoop, S. Raghavendra
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/11027096/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849336932900798464
author K. Samanth
Ramyashree Ramyashree
B. N. Anoop
S. Raghavendra
author_facet K. Samanth
Ramyashree Ramyashree
B. N. Anoop
S. Raghavendra
author_sort K. Samanth
collection DOAJ
description The escalating crisis of marine pollution, especially the buildup of underwater waste, poses a significant threat to ocean ecosystems and marine life. Identifying and removing submerged debris is challenging due to poor visibility, complex water conditions, and the inefficiency of traditional manual approaches. This study leverages deep learning-based object detection techniques to improve underwater waste identification, focusing on the YOLOv8 and YOLOv9 architectures. Using the TrashCan 1.0 dataset—a large collection of labeled underwater images—the models were trained and evaluated based on key metrics such as mean Average Precision (mAP) and Intersection over Union (IoU). Results show that the YOLOv9c model performed best on the 3-Class dataset, achieving a mAP50:95 of 0.759, a mAP50 of 0.934, precision of 0.915, and recall of 0.877, highlighting its high accuracy and robustness. Meanwhile, YOLOv8l showed slightly better results on the 4-Class dataset with a mAP50:95 of 0.725, indicating dataset-dependent performance variations. The ability of these models to deliver real-time detection makes them ideal for integration into autonomous underwater vehicles (AUVs) and remotely operated vehicles (ROVs), enabling large-scale monitoring and cleanup operations. By advancing AI-driven solutions for marine conservation, this research contributes to more efficient and scalable efforts in mitigating the long-term impact of ocean pollution.
format Article
id doaj-art-37c5aa491d3545a8a0c0d7e5890dc1d3
institution Kabale University
issn 2169-3536
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-37c5aa491d3545a8a0c0d7e5890dc1d32025-08-20T03:44:51ZengIEEEIEEE Access2169-35362025-01-0113994469946410.1109/ACCESS.2025.357723911027096A Comprehensive Study On Underwater Object Detection Using Deep Neural NetworksK. Samanth0https://orcid.org/0009-0004-3085-8910Ramyashree Ramyashree1https://orcid.org/0000-0002-0237-2444B. N. Anoop2https://orcid.org/0000-0002-6082-391XS. Raghavendra3https://orcid.org/0000-0003-2733-3916Department of Information and Communication Technology, Manipal Institute of Technology, Manipal Academy of Higher Education, Manipal, IndiaDepartment of Information and Communication Technology, Manipal Institute of Technology, Manipal Academy of Higher Education, Manipal, IndiaDepartment of Information and Communication Technology, Manipal Institute of Technology, Manipal Academy of Higher Education, Manipal, IndiaDepartment of Information and Communication Technology, Manipal Institute of Technology, Manipal Academy of Higher Education, Manipal, IndiaThe escalating crisis of marine pollution, especially the buildup of underwater waste, poses a significant threat to ocean ecosystems and marine life. Identifying and removing submerged debris is challenging due to poor visibility, complex water conditions, and the inefficiency of traditional manual approaches. This study leverages deep learning-based object detection techniques to improve underwater waste identification, focusing on the YOLOv8 and YOLOv9 architectures. Using the TrashCan 1.0 dataset—a large collection of labeled underwater images—the models were trained and evaluated based on key metrics such as mean Average Precision (mAP) and Intersection over Union (IoU). Results show that the YOLOv9c model performed best on the 3-Class dataset, achieving a mAP50:95 of 0.759, a mAP50 of 0.934, precision of 0.915, and recall of 0.877, highlighting its high accuracy and robustness. Meanwhile, YOLOv8l showed slightly better results on the 4-Class dataset with a mAP50:95 of 0.725, indicating dataset-dependent performance variations. The ability of these models to deliver real-time detection makes them ideal for integration into autonomous underwater vehicles (AUVs) and remotely operated vehicles (ROVs), enabling large-scale monitoring and cleanup operations. By advancing AI-driven solutions for marine conservation, this research contributes to more efficient and scalable efforts in mitigating the long-term impact of ocean pollution.https://ieeexplore.ieee.org/document/11027096/Autonomous underwater vehiclesdeep learningmarine waste detectionunderwater object detectionYOLOv8YOLOv9
spellingShingle K. Samanth
Ramyashree Ramyashree
B. N. Anoop
S. Raghavendra
A Comprehensive Study On Underwater Object Detection Using Deep Neural Networks
IEEE Access
Autonomous underwater vehicles
deep learning
marine waste detection
underwater object detection
YOLOv8
YOLOv9
title A Comprehensive Study On Underwater Object Detection Using Deep Neural Networks
title_full A Comprehensive Study On Underwater Object Detection Using Deep Neural Networks
title_fullStr A Comprehensive Study On Underwater Object Detection Using Deep Neural Networks
title_full_unstemmed A Comprehensive Study On Underwater Object Detection Using Deep Neural Networks
title_short A Comprehensive Study On Underwater Object Detection Using Deep Neural Networks
title_sort comprehensive study on underwater object detection using deep neural networks
topic Autonomous underwater vehicles
deep learning
marine waste detection
underwater object detection
YOLOv8
YOLOv9
url https://ieeexplore.ieee.org/document/11027096/
work_keys_str_mv AT ksamanth acomprehensivestudyonunderwaterobjectdetectionusingdeepneuralnetworks
AT ramyashreeramyashree acomprehensivestudyonunderwaterobjectdetectionusingdeepneuralnetworks
AT bnanoop acomprehensivestudyonunderwaterobjectdetectionusingdeepneuralnetworks
AT sraghavendra acomprehensivestudyonunderwaterobjectdetectionusingdeepneuralnetworks
AT ksamanth comprehensivestudyonunderwaterobjectdetectionusingdeepneuralnetworks
AT ramyashreeramyashree comprehensivestudyonunderwaterobjectdetectionusingdeepneuralnetworks
AT bnanoop comprehensivestudyonunderwaterobjectdetectionusingdeepneuralnetworks
AT sraghavendra comprehensivestudyonunderwaterobjectdetectionusingdeepneuralnetworks