kfas

SEARCH FOR ARTICLE

kfas, vol. 57, no. 5, pp.581-588, October, 2024 DOI. https://doi.org/10.5657/KFAS.2024.0581

Machine Learning-based Data Construction and Model Evaluation for Monitoring of Giant Jellyfish Nemopilema nomurai

오선영·김형태·이경훈1*
국립부경대학교 해양생산관리학부 수산물리학전공, 1국립부경대학교 해양생산시스템관리학부

ABSTRACT

In this, we study developed a machine-learning system that can effectively detect giant jellyfish Nemopilema nomurai by collecting videos of their appearances. Surveys were conducted in the East China Sea, South Sea, and Jeju coastal waters, which are presumed to be jellyfish migration routes. Video data were collected using GoPro cameras, and images were extracted at 1 fps to train the YOLOv8 Nano and Medium models. The YOLOv8 Nano model achieved an F1 score of 0.83 with high confidence and maintained high precision in the precision-recall curve, demonstrating its effectiveness in predicting jellyfish occurrences. The YOLOv8 nano model demonstrated excellent reliability and precision, indicating its potential for effective jellyfish detection. However, to improve the performance of the model even further, data from various environments must be collected and additional validations must be performed.
Keyword

Giant jellyfish, Machine learning, YOLOv8, Precision-recall curve