AI-Powered Tracking for Sustainable Marine Ecosystem Resource Management Projects

Abstract

Ecosystems are our planet's life-support systems that facilitate sustainable development. Within the marine ecosystem, oysters serve as a keystone species. Numerous oyster restoration projects have been launched with a crucial element involving precise assessment of oyster population sizes within specific reef areas. However, the current methods of tracking oyster populations are approximate and lack precision. To address this research gap, the authors developed an AI-empowered project for oyster detection. Specifically, they created a dataset of wild oysters, utilized Roboflow for image annotation, and employed image augmentation techniques to augment the training data. Then, they fine-tuned a YOLOv8 computer vision object detection model using their dataset. The results demonstrated a mean average precision (mAP) of 85.2 percent and an accuracy of 87.7 percent for oyster detection. This approach improved upon previous attempts to detect wild oysters, offering a more effective solution for population assessment, which is a fundamental step toward sustainable oyster restoration project management.