Manuscript received September 2, 2024; revised September 20, 2024; accepted November 11, 2024; published May 20, 2025.

Abstract—This study investigates the extent to which the YOLOv9e-instance segmentation model classifies and detects different types of marine objects, such as corals, marine life, seagrass, and seaweed. This study utilizes image augmentation techniques to improve the detection and classification of objects using YOLOv9. The study emphasizes the need to examine the distribution of classes within the dataset, as class imbalances can have a major impact on the model’s performance. Throughout the training, the model showed a constant decrease in loss functions such as box loss, segmentation loss, and classification loss, demonstrating effective learning and generalization. The precision and recall metrics improved significantly, with a mean Average Precision (mAP) of 0.883 at an Intersectionover- Union (IoU) threshold of 0.5, validating the model’s high accuracy across classes. The F1-Confidence Curve study yielded an overall F1 score of 0.84 at a confidence threshold of 0.534, highlighting the model’s robustness in achieving a balance between precision and recall. The results suggest that while the model excels in detecting corals, seagrass, and seaweed, it faces challenges in accurately identifying marine life, pointing to the need for additional refinement to address class imbalances.

Keywords—YOLOv9, instance segmentation

Cite: Ken D. Gorro, Anthony S. Ilano, Lawrence P. Roble, Rue Nicole R. Santillan, Joseph C. Pepito, Elmo B. Ranolo, Kim D. Gorro, Apple Jane M. Gorro, Moustafa F. Ali, Archival J. Sebial, and Jezreel N. Buot, "Detection of Corals, Seagrass, and Seaweeds Using YOLOv9 Instance Segmentation with Image Augmentation," Journal of Image and Graphics, Vol. 13, No. 3, pp. 231-244, 2025.

Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC-BY-4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.